diff --git a/.gitignore b/.gitignore
index 5b59591..63df802 100644
--- a/.gitignore
+++ b/.gitignore
@@ -4,3 +4,4 @@ playwright-report
 test-results
 coverage
 docs
+references/.venv
diff --git a/JSDOC_README.md b/JSDOC_README.md
index 4f9b9c0..d93706f 100644
--- a/JSDOC_README.md
+++ b/JSDOC_README.md
@@ -1,6 +1,6 @@
-# Ghostati - Technical Documentation
+# Ghostmaxxing - Technical Documentation
 
-Welcome to the internal technical documentation for Ghostati, a Web AR laboratory for the development and real-time testing of anti-biometric facial recognition camouflage (CV Dazzle).
+Welcome to the internal technical documentation for Ghostmaxxing, a Web AR laboratory for the development and real-time testing of anti-biometric facial recognition camouflage (CV Dazzle).
 
 This documentation is generated from the source code using JSDoc. It provides a detailed breakdown of the functions, classes, and namespaces used throughout the project. 
 
diff --git a/PROMPT-REFERENCES-UPDATE.txt b/PROMPT-REFERENCES-UPDATE.txt
deleted file mode 100644
index 7ef1977..0000000
--- a/PROMPT-REFERENCES-UPDATE.txt
+++ /dev/null
@@ -1,86 +0,0 @@
-Produce or extend a JSON reference set for the same project universe as the existing file "REFERENCES.json".
-
-Goal:
-Maintain a coherent, deduplicated set of cultural, artistic, academic, and activism-adjacent references around:
-- anti-surveillance aesthetics
-- face obfuscation / face camouflage
-- makeup-based or appearance-based interventions
-- adversarial attacks on face recognition
-- adjacent physical-world adversarial vision techniques
-
-Output rules:
-1. Output valid JSON only.
-2. Preserve the top-level structure:
-{
-  "schema_version": "2.0",
-  "project_scope": "...",
-  "ordering_definition": "...",
-  "closeness_definition": {...},
-  "preview_image_definition": "...",
-  "references": [...]
-}
-3. Each reference object MUST contain at least:
-- slug
-- title
-- author
-- year
-- link
-- type
-- closeness
-- description
-- preview_image
-
-Field conventions:
-- slug: kebab-case unique identifier stable across updates.
-- title: canonical public title of the work/project/paper.
-- author: array of author/creator names in display order.
-- year: integer year of publication, release, first public presentation, or first documented appearance.
-- link: stable canonical link preferred in this order:
-  DOI > arXiv abs page > official project page > official institutional page.
-- type: exactly one of "artistic" | "research" | "activism".
-- closeness: integer from 1 to 100, not a string and no percent sign.
-- description: short editorial description in Italian, ideally 1–2 sentences, written for exhibition/timeline display.
-- preview_image: local site-root path beginning with / and pointing to an image asset, e.g. "/references-v2-previews/my-item.svg".
-
-Closeness rubric:
-- 100 = exact anchor cluster for this project world:
-  * Michelle Tylicki + Lauri Love — DAZZLE
-  * Adam Harvey — CV Dazzle
-- 90–99 = direct face-centered anti-surveillance / makeup / camouflage references.
-- 70–89 = directly relevant face-recognition attack literature or highly adjacent face-obfuscation work.
-- 40–69 = enabling or adjacent adversarial-vision methods that matter conceptually but are less face/makeup-specific.
-- 1–39 = peripheral or modality-shifted references, especially infrared/night-vision/sensor-jamming examples.
-
-Preview-image rules:
-- preview_image should point to a local asset, not a remote URL.
-- If no authoritative image is available, use a typological illustration or icon that represents the strategy visually.
-- Prefer one preview image per entry, stable over time.
-- Do not leave preview_image empty.
-
-Normalization rules:
-- Deduplicate by normalized title first, then by DOI or canonical link.
-- Prefer canonical capitalization.
-- Prefer full author lists when available.
-- Do not invent missing metadata.
-- If metadata is uncertain, keep the source out rather than guessing.
-- Keep the existing entry if the new candidate is the same work with a worse link.
-- If the same work exists with a better canonical link, replace only the link and preserve closeness unless there is a strong reason to revise it.
-- Preserve the existing slug whenever the same work reappears.
-
-Ordering logic:
-- Return the full updated JSON sorted newest first:
-  year descending, closeness descending, title ascending.
-
-Style constraints:
-- Avoid generic press coverage unless it is itself the primary artistic or activist source.
-- Prefer official project pages, DOI links, arXiv abs pages, institutional pages, or artist pages.
-- Keep the set operational: every added item should help situate the project culturally, artistically, academically, or politically.
-- Descriptions should be concise, readable, and suitable for a static timeline in Hugo.
-
-When adding new references:
-- Merge them into the existing JSON instead of rewriting from scratch.
-- Preserve the schema_version, ordering_definition, closeness_definition, and preview_image_definition unless explicitly asked to change them.
-- Maintain internal consistency with previously assigned closeness values.
-
-If asked to append items, return the full updated JSON, not just the delta.
-
diff --git a/README.it.md b/README.it.md
deleted file mode 100644
index 1090e9c..0000000
--- a/README.it.md
+++ /dev/null
@@ -1,83 +0,0 @@
-# Ghòstati | Face Lab
-
-Laboratorio di trucco avversario\* per ingannare il riconoscimento facciale agendo su pochi punti del tuo volto.
-
-
-\* Come tradurreste voi "adversarial makeup" ? trucco antagonista? trucco avversariale? 
-
-## Ispirato alle ricerche di (References)
-
-- [DAZZLE](https://www.michelletylicki.info/dazzle/) — Michelle Tylicki, Lauri Love (2023). Camouflage facciale come gesto estetico-politico contro la sorveglianza biometrica. *artistico*
-- [CV Dazzle](https://adam.harvey.studio/cvdazzle/) — Adam Harvey (2010). Metodo iconico di trucco/hairstyle per disturbare il rilevamento facciale automatico. *artistico*
-- [The Dazzle Club](https://emilyroderick.com/work/the-dazzle-club/) — Evie Price, Emily Roderick, Georgina Rowlands, Anna Hart (2019). Azioni urbane contro il riconoscimento facciale con mascheramento creativo. *artistico*
-- [Adv-Makeup: A New Imperceptible and Transferable Attack on Face Recognition](https://arxiv.org/abs/2105.03162) — Bangjie Yin et al. (2021). Attacco makeup trasferibile e poco percettibile contro sistemi di face recognition. *pubblicazione*
-- [Accessorize to a Crime: Real and Stealthy Attacks on State-of-the-Art Face Recognition](https://doi.org/10.1145/2976749.2978392) — Mahmood Sharif, Sruti Bhagavatula, Lujo Bauer, Michael K. Reiter (2016). Attacchi fisici stealth con accessori contro modelli di riconoscimento facciale. *pubblicazione*
-- [HyperFace](https://adam.harvey.studio/hyperface/) — Adam Harvey (2016). Pattern tessili che massimizzano falsi positivi dei detector facciali. *artistico*
-- [Adversarial Generative Nets: Neural Network Attacks on State-of-the-Art Face Recognition](https://arxiv.org/abs/1801.00349) — Mahmood Sharif, Sruti Bhagavatula, Lujo Bauer, Michael K. Reiter (2017). Approccio generativo per perturbare in modo mirato il riconoscimento del volto. *pubblicazione*
-- [Adversarial Attacks against Face Recognition: A Comprehensive Study](https://arxiv.org/abs/2007.11709) — Fatemeh Vakhshiteh, Ahmad Nickabadi, Raghavendra Ramachandra (2020). Survey completo su tecniche di attacco ai sistemi di riconoscimento facciale. *pubblicazione*
-- [VLA: A Practical Visible Light-based Attack on Face Recognition Systems in Physical World](https://doi.org/10.1145/3351261) — Meng Shen, Zelin Liao, Liehuang Zhu, Ke Xu, Xiaojiang Du (2019). Attacco fisico con luce visibile per degradare l'identificazione facciale. *pubblicazione*
-- [Adversarial Robustness Toolbox v1.0.0](https://arxiv.org/abs/1807.01069) — Maria-Irina Nicolae et al. (2018). Libreria per valutare robustezza e attacchi avversari in pipeline ML. *altro*
-- [Adversarial Patch](https://arxiv.org/abs/1712.09665) — Tom B. Brown, Dandelion Mané, Aurko Roy, Martín Abadi, Justin Gilmer (2017). Patch fisiche stampabili che causano errori di classificazione robusti. *pubblicazione*
-- [Adversarial Manipulation of Deep Representations](https://arxiv.org/abs/1511.05122) — Sara Sabour, Yanshuai Cao, Fartash Faghri, David J. Fleet (2015). Manipolazione di rappresentazioni profonde per obiettivi avversari. *pubblicazione*
-- [DPatch: An Adversarial Patch Attack on Object Detectors](https://arxiv.org/abs/1806.02299) — Xin Liu, Huanrui Yang, Ziwei Liu, Linghao Song, Hai Li, Yiran Chen (2018). Patch avversaria fisica per compromettere object detector in scena reale. *pubblicazione*
-- [ShapeShifter: Robust Physical Adversarial Attack on Faster R-CNN Object Detector](https://arxiv.org/abs/1804.05810) — Shang-Tse Chen, Cory Cornelius, Jason Martin, Duen Horng Chau (2018). Esempi fisici robusti contro detector basati su Faster R-CNN. *pubblicazione*
-- [Breaking certified defenses: Semantic adversarial examples with spoofed robustness certificates](https://arxiv.org/abs/2003.08937) — Amin Ghiasi, Ali Shafahi, Tom Goldstein (2020). Dimostra limiti delle difese certificate con esempi avversari semantici. *pubblicazione*
-- [Physical-World Optical Adversarial Attacks on 3D Face Recognition](https://arxiv.org/abs/2205.13412) — Yanjie Li, Yiquan Li, Xuelong Dai, Songtao Guo, Bin Xiao (2022). Attacchi ottici nel mondo fisico contro riconoscimento facciale 3D. *pubblicazione*
-- [Human-Imperceptible Physical Adversarial Attack for NIR Face Recognition Models](https://arxiv.org/abs/2504.15823) — Songyan Xie, Jinghang Wen, Encheng Su, Qiucheng Yu (2025). Attacco fisico quasi impercettibile su modelli NIR per il volto. *pubblicazione*
-- [Accessorize in the Dark: A Security Analysis of Near-Infrared Face Recognition](https://doi.org/10.1007/978-3-031-51479-1_3) — Amit Cohen, Mahmood Sharif (2024). Analisi di sicurezza su riconoscimento facciale near-infrared e bypass pratici. *pubblicazione*
-- [The Camera Shy Hoodie](https://www.macpierce.com/the-camera-shy-hoodie) — Mac Pierce (2023). Capo wearable pensato per disturbare cattura e analisi visiva. *artistico*
-
-# Ecco a voi: Ghòstati!
-
-![Ghòstati](facerec-transparency.png)
-
-## Panoramica
-
-**Ghòstati** è una piattaforma sperimentale e uno strumento diagnostico progettato per contrastare gli algoritmi di riconoscimento facciale. Applicando specifici pattern di trucco (ispirati al concetto di CV Dazzle), gli utenti possono esplorare come i modelli di computer vision interpretano i landmark facciali e tentare di offuscare la propria identità digitale in tempo reale.
-
-Il progetto presenta un'architettura modulare basata su plugin, la quale permette a qualsiasi sviluppatore di scrivere script di trucco AR personalizzati ("Ghostyles") e di testarne l'efficacia contro i modelli di riconoscimento direttamente nel browser tramite la webcam.
-
-## Funzionalità Principali
-
-- **Live Face Tracking:** Rilevamento dei landmark facciali in tempo reale direttamente nel browser utilizzando `face-api.js`.
-- **Sistema di Plugin Modulare (Ghostyles):** Carica dinamicamente effetti di trucco AR personalizzati. I plugin possono essere ospitati localmente o tramite URL remoto. Alcuni effetti basilari (che non ostacolano il riconoscimento ma aiutano a capire il funzionamento del codice) sono :
-  - Graphic Liner, Smokey Eyes, Blush Lift, Lip Tint, Soft Contour, Stage Mask, Splash, etc.
-  - La pagina di documentazione dei plug-in: https://sindacato.nina.watch/ghostati/ghostati-docs.html
-- **Modalità Diagnostica ("Scansione Trucco"):** Testa l'efficacia del tuo camouflage AR. Lo strumento valuta l'opacità del trucco, cattura il volto alterato e calcola la probabilità di corrispondenza rispetto ai profili salvati per determinare se il sistema di riconoscimento è stato ingannato.
-- **Salva e Confronta (Enrolling):** Salva un volto di base iniziale e confrontalo con il feed live della webcam per verificare se l'algoritmo di face matching ti riconosce ancora dopo aver applicato il camuffamento.
-- **Privacy-First:** Tutte le elaborazioni vengono eseguite localmente sul computer, senza caricare dati biometrici su server remoti.
-
-## Installazione
-
-Trattandosi di un'applicazione web statica, non è necessario alcun passaggio di "build", e si può testare online a [https://sindacato.nina.watch/ghostati](https://sindacato.nina.watch/ghostati).
-
-1. Clona il repository:
-   ```bash
-   git clone https://github.com/vecna/ghostati.git
-   cd ghostati
-   ```
-2. Avvia un server web HTTP locale nella cartella:
-   ```bash
-   npx http-server . 
-   # oppure
-   python3 -m http.server 8000
-   ```
-3. Apri un browser moderno e vai all'indirizzo `http://localhost:8000/ghostati-face-api.html`.
-
-## Sviluppare un Ghostyle (Plugin)
-
-Puoi creare i tuoi effetti modulari di trucco AR chiamati **Ghostyles**. Un "Ghostyle" è un semplice modulo JavaScript che esporta una funzione di disegno atta ad agganciarsi al motore di face tracking.
-
-Per sviluppare un nuovo Ghostyle:
-1. Fai una copia di `./ghostyles/00-template.js`.
-2. Implementa la tua logica su canvas basandoti sui landmark facciali forniti ad ogni frame.
-3. Testalo dal vivo incollando l'URL locale/remoto nel box "Carica Ghostyle Remoto" nella pagina!
-
-Consulta la pagina `ghostati-docs.html` per una documentazione più avanzata sullo sviluppo dei Ghostyles.
-
-## *Contesto a Maggio 2026*
-
-Presentato all'interno del **Festival di NINA**, questo strumento mira a sensibilizzare l'opinione pubblica riguardo la sorveglianza biometrica e l'uso delle tecnologie di riconoscimento facciale.
-
----
-*Per la versione in inglese, consulta [README.md](README.md).*
diff --git a/README.md b/README.md
index 72ab975..cbbc3f8 100644
--- a/README.md
+++ b/README.md
@@ -1,156 +1,330 @@
-## Status
-![Unit Test Coverage](https://img.shields.io/badge/coverage-52.02%25-lightgrey)
-# ghòstati! | the _Face Confusion Assistant?_
-
-This is a: Web AR laboratory for the development and real-time testing of anti-biometric facial recognition camouflage (also known as, Computer Vision Dazzle).
-
-![ghòstati](facerec-transparency.png)
-
-## Overview
-
-**ghòstati** is an experimental platform and diagnostic tool designed to counter facial recognition algorithms. By applying specific makeup patterns (inspired by the CV Dazzle concept), users can explore how computer vision models interpret facial landmarks and attempt to anonymize their digital footprint in real time. 
-
-The project features a fully modular, plugin-based architecture, allowing any developer to write custom AR makeup scripts ("Ghostyles") and test their efficiacy against recognition models directly in the browser via their webcam.
-
-
-
-## Features
-- **Live Face Tracking:** Real-time facial landmark detection directly in the browser utilizing `face-api.js`.
-- **Modular Plugin System (Ghostyles):** Load custom AR makeup effects dynamically. Plugins can be hosted locally or loaded via a remote URL. Included effects: 
-  - Graphic Liner, Smokey Eyes, Blush Lift, Lip Tint, Soft Contour, Stage Mask, Splash, etc.
-- **Diagnostic Mode ("Scansione Trucco"):** Test the effectiveness of your AR camouflage. The tool evaluates makeup opacity, captures the altered face, and computes matching likelihood against saved profiles to determine if the face recognition system is successfully spoofed.
-- **Save & Compare:** Save an initial baseline face and compare live webcam feeds to it to check if the face matching algorithm still recognizes you after applying the camouflage.
-- **Privacy-First:** All processing is done locally on the client interface without uploading biometric data to remote servers.
-
-## Getting Started
-
-Since it's a static web application, there is no build step required.
-
-1. Clone the repository:
-   ```bash
-   git clone https://github.com/vecna/ghostati.git
-   cd ghostati
-   ```
-2. Serve the directory with a local HTTP server:
-   ```bash
-   npx http-server . 
-   # or
-   python3 -m http.server 8000
-   ```
-3. Open a modern browser, ensure to have a webcam, and navigate to `http://localhost:8000/ghostati.html`.
-
-## Generating Documentation
-
-This project uses [JSDoc](https://jsdoc.app/) to generate documentation for the source code.
-
-To generate the documentation, run the following command:
-```bash
-npm run docs
+# ghòstati
+
+```text
+        _          _        _   _
+   __ _| |__   ___| |_ __ _| |_(_)
+  / _` | '_ \ / _ \ __/ _` | __| |
+ | (_| | | | | (_) | || (_| | |_| |
+  \__, |_| |_|\___/ \__\__,_|\__|_|
+  |___/     Web AR face-recognition test lab
 ```
-This will parse the JSDoc comments in the `scripts/` directory and generate a static HTML website in the `docs/` folder. You can open `docs/index.html` in your browser to view the documentation.
-
-## Writing a Ghostyle (Plugin)
-
-You can create your own modular AR makeup effects called **Ghostyles**. A "Ghostyle" is a simple JavaScript module that exports a draw function hooking into the face tracking engine. 
-
-To develop a new Ghostyle:
-1. Copy the `./ghostyles/00-template.js`.
-2. Implement your custom canvas drawing logic based on the provided facial landmarks.
-3. Test it live by editing the `ghostylist.json` file
-4. Open a PR to get it distributed - this can't get any easy at the moment.
-
-See the `ghostati-docs.html` page for more advanced documentation on Ghostyle development.
-
-## Context
-
-Presented as part of the **NINA Festival**, this tool aims to raise awareness regarding biometric surveillance and facial recognition technologies.
-
----
-*For the Italian version, please see [README.it.md](README.it.md).*
-
-
-
-
-
-
-
-
-
 
+[![Unit Test Coverage](https://img.shields.io/badge/coverage-53.45%25-lightgrey)](coverage/)
+[![Docs](https://img.shields.io/badge/docs-JSDoc-blue)](docs/)
+[![Source](https://img.shields.io/badge/source-GitHub-black)](https://github.com/vecna/ghostati)
 
+**ghòstati** is a static, browser-side Web AR laboratory for designing and testing face-obfuscation overlays against face-detection and face-recognition pipelines.
 
+The project combines webcam capture, [`face-api.js`](https://github.com/vladmandic/face-api), [MediaPipe Tasks Vision](https://developers.google.com/mediapipe/solutions/vision/face_landmarker), canvas rendering, local descriptor storage, and a plugin system for custom **Ghostyles**: 2D or 3D face overlays that can be tested against recognition behavior in real time.
 
+Primary links:
 
+- Live project root: [https://sindacato.nina.watch/ghostati/](https://sindacato.nina.watch/ghostati/)
+- Browser app: [ghostati.html](https://sindacato.nina.watch/ghostati/ghostati.html)
+- Source code: [github.com/vecna/ghostati](https://github.com/vecna/ghostati)
+- Ghostyle gallery / distribution site: [ghostyles.vecna.eu](https://ghostyles.vecna.eu)
+- Generated API docs: [docs/](https://sindacato.nina.watch/ghostati/docs/)
+- Legacy docs page: [ghostati-docs.html](https://sindacato.nina.watch/ghostati/ghostati-docs.html)
+- Test coverage: [coverage/](https://sindacato.nina.watch/ghostati/coverage/)
+- Project context page: [sindacato.nina.watch/it/iniziative/ghostati](https://sindacato.nina.watch/it/iniziative/ghostati/)
+- Sitemap: [sitemap.xml](https://sindacato.nina.watch/ghostati/sitemap.xml)
+- Reference dataset: [REFERENCES.json](REFERENCES.json)
+- Reference-update prompt: [PROMPT-REFERENCES-UPDATE.txt](PROMPT-REFERENCES-UPDATE.txt)
 
+**Central field-reporting resource:** if you know of a place where facial recognition is being deployed, tested, procured, or hidden in public-space infrastructure, use the NINA submission node: [Raccontacelo](https://raccontaci.nina.watch/#/submission?context=10c78596-3ea0-4867-b2fb-21fdb8e3f40c). Reports about supplier, technology, data access, deployment context, limits, and abuses are project inputs, not side notes.
 
+# What the app does
 
+```text
+ webcam ──► detector ──► landmarks ──► overlay renderer
+    │                         │                │
+    └──── baseline face DB ◄──┴──── compare ◄──┘
+```
 
+The app is designed around a simple experimental loop:
+
+1. Start the webcam in a modern browser.
+2. Load face-detection and landmark models.
+3. Save a local baseline descriptor for a consenting test face.
+4. Apply a Ghostyle overlay to the live video/canvas layer.
+5. Re-run detection and recognition against the saved descriptor.
+6. Observe whether the pipeline still detects the face, extracts landmarks, and matches the baseline.
+
+Core capabilities:
+
+- live webcam setup and teardown through [`scripts/camera.js`](scripts/camera.js);
+- face detection, landmarks, descriptors, and match orchestration through [`scripts/engine.js`](scripts/engine.js);
+- 3D/MediaPipe loop support through [`scripts/mediapipe-loop.js`](scripts/mediapipe-loop.js) and [`scripts/engine-3d.js`](scripts/engine-3d.js);
+- bounding-box overlays through [`scripts/bbox-overlay.js`](scripts/bbox-overlay.js);
+- dynamic Ghostyle loading through [`scripts/ghostyles-manager.js`](scripts/ghostyles-manager.js);
+- 3D plugin loading through [`scripts/plugins3d-loader.js`](scripts/plugins3d-loader.js);
+- IndexedDB-backed local state through [`scripts/db.js`](scripts/db.js);
+- DOM and UI bindings through [`scripts/dom.js`](scripts/dom.js), [`scripts/main.js`](scripts/main.js), and [`scripts/ghostati-mobile-ui.js`](scripts/ghostati-mobile-ui.js);
+- image/makeup export helpers through [`scripts/export-makeup.js`](scripts/export-makeup.js);
+- landing-page animation through [`scripts/index-effect.js`](scripts/index-effect.js).
+
+# Runtime architecture
+
+```text
+ ghostati.html
+   ├─ @vladmandic/face-api
+   ├─ @mediapipe/tasks-vision
+   ├─ scripts/main.js
+   │   ├─ camera.js
+   │   ├─ engine.js
+   │   ├─ db.js
+   │   ├─ dom.js
+   │   └─ ghostyles-manager.js
+   ├─ ghostyles.json
+   └─ ghostyles/*.js
+```
 
+The project is a static web app: there is no production build step required to open the interface locally. The browser loads HTML, CSS, JavaScript modules, model assets, and plugin manifests.
 
+Important local entry points:
 
+- [`index.html`](index.html) — public landing page with links to code, docs, coverage, Ghostyles, project context, and the reporting node.
+- [`ghostati.html`](ghostati.html) — main webcam/AR application.
+- [`ghostyles.json`](ghostyles.json) — Ghostyle manifest.
+- [`JSDOC_README.md`](JSDOC_README.md) — concise generated-docs overview.
+- [`REFERENCES.json`](REFERENCES.json) — curated technical/cultural reference set.
 
+Runtime external dependencies visible from the HTML/config layer:
 
+- [Google Fonts](https://fonts.googleapis.com) / [Google Fonts static assets](https://fonts.gstatic.com)
+- [Landing-page Google Fonts CSS](https://fonts.googleapis.com/css2?family=League+Script&family=Outfit:wght@400;600;700;800;900&family=JetBrains+Mono:wght@400;700&display=swap)
+- [jsDelivr CDN](https://cdn.jsdelivr.net)
+- [`@vladmandic/face-api`](https://cdn.jsdelivr.net/npm/@vladmandic/face-api/dist/face-api.js)
+- [face-api.js model weights](https://cdn.jsdelivr.net/gh/justadudewhohacks/face-api.js-models@master/)
+- [MediaPipe Tasks Vision](https://cdn.jsdelivr.net/npm/@mediapipe/tasks-vision@0.10.35)
+- [MediaPipe Face Landmarker model](https://storage.googleapis.com/mediapipe-models/face_landmarker/face_landmarker/float16/1/face_landmarker.task)
+- [MediaPipe Image Embedder model](https://storage.googleapis.com/mediapipe-models/image_embedder/mobilenet_v3_small/float32/1/mobilenet_v3_small.tflite)
 
+For workshops or higher-risk demos, prefer self-hosting model and library assets instead of relying on third-party CDNs.
 
+# Install and run locally
 
+```text
+ clone ──► install dev deps ──► static server ──► browser + webcam
+```
 
-**Last commit:** `8296992` – added the UX content for ghostyles future developer
-
-
-
-
-
-
-
-
-
-
-
+Clone the repository:
 
+```bash
+git clone https://github.com/vecna/ghostati.git
+cd ghostati
+```
 
+Install development dependencies:
 
+```bash
+npm install
+```
 
+Serve the directory with any local static server:
 
+```bash
+npx http-server .
+# or
+python3 -m http.server 8000
+```
 
+Open the app:
 
+```text
+http://localhost:8000/ghostati.html
+```
 
+Open the landing page:
 
+```text
+http://localhost:8000/
+```
 
+# Ghostyles plugin API
 
+```text
+ landmarks + box + canvas context
+          │
+          ▼
+      ghostyle module
+          │
+          ▼
+   live overlay + diagnostic pass
+```
 
+A **Ghostyle** is a JavaScript module that draws an overlay anchored to a detected face. It can be local or loaded through a manifest.
+
+Start from [`ghostyles/00-template.js`](ghostyles/00-template.js), then add the file to [`ghostylist.json`](ghostylist.json). Existing 2D examples include:
+
+- [`ghostyles/graphic-liner.js`](ghostyles/graphic-liner.js)
+- [`ghostyles/smokey-eyes.js`](ghostyles/smokey-eyes.js)
+- [`ghostyles/blush-lift.js`](ghostyles/blush-lift.js)
+- [`ghostyles/lip-tint.js`](ghostyles/lip-tint.js)
+- [`ghostyles/soft-contour.js`](ghostyles/soft-contour.js)
+- [`ghostyles/stage-mask.js`](ghostyles/stage-mask.js)
+- [`ghostyles/splash.js`](ghostyles/splash.js)
+
+A minimal 2D plugin shape:
+
+```js
+/**
+ * @name Example Ghostyle
+ * @engine faceapi
+ */
+export function onInit() {
+  return 'loaded';
+}
+
+export function onDraw(ctx, landmarks, box) {
+  ctx.save();
+  // Draw against landmarks and detection box here.
+  ctx.restore();
+}
+
+export function onClear(ctx) {
+  ctx.clearRect(0, 0, ctx.canvas.width, ctx.canvas.height);
+}
+```
 
+3D/MediaPipe-oriented examples live in:
 
+- [`ghostyles3d/prove-stripes.js`](ghostyles3d/prove-stripes.js)
+- [`ghostyles3d/uv-stripes.js`](ghostyles3d/uv-stripes.js)
 
+Register them in [`ghostylist3d.json`](ghostylist3d.json).
 
+# Testing and generated docs
 
+```text
+ unit tests ──► coverage
+ e2e tests  ──► browser flows
+ jsdoc      ──► docs/
+```
 
+Package scripts:
 
+```bash
+npm run test:unit
+npm run test:unit -- --coverage
+npm run test:e2e
+npm run docs
+```
 
+Testing stack:
 
+- [Vitest](https://vitest.dev/) for unit tests;
+- [`@vitest/coverage-v8`](https://vitest.dev/guide/coverage) for coverage;
+- [Playwright](https://playwright.dev/) for browser-level tests;
+- [JSDOM](https://github.com/jsdom/jsdom) and [node-canvas](https://github.com/Automattic/node-canvas) for DOM/canvas test fixtures;
+- [JSDoc](https://jsdoc.app/) with [`clean-jsdoc-theme`](https://github.com/ankitskvmdam/clean-jsdoc-theme) for generated documentation.
 
+Relevant test directories:
 
+- [`tests/unit/`](tests/unit/)
+- [`tests/e2e/`](tests/e2e/)
+- [`tests/fixtures/`](tests/fixtures/)
 
+# Privacy, data, and limitations
 
+```text
+ browser storage
+    ├─ descriptors
+    ├─ preferences
+    └─ test state
+ external network
+    ├─ fonts
+    ├─ CDN libraries
+    └─ model weights
+```
 
+The app is designed to keep biometric test data local to the browser interface. Face descriptors and related state are stored locally, not posted to a central server by the default app flow.
 
+Technical caveats:
 
+- webcam access is controlled by the browser permission model;
+- local descriptors may persist in IndexedDB/local browser storage until cleared;
+- screenshots, recordings, and exports should be treated as sensitive biometric-adjacent material;
+- CDN-loaded libraries and model files still create external network requests;
+- detection failure, landmark instability, and match failure are different outcomes and should not be collapsed into “anonymity”;
+- a Ghostyle that affects this browser pipeline may not affect another face-recognition system.
 
+Use consenting test subjects. Do not represent experimental overlays as operational safety guarantees.
 
+# Field reports: public-space face recognition
 
+```text
+ observe ──► document ──► submit ──► update resistance research
+```
 
+Ghostmaxxing is also connected to a field-reporting workflow. The landing page now treats the NINA reporting node as a central project resource:
 
+[Raccontacelo: segnala un possibile uso di riconoscimento facciale nello spazio pubblico](https://raccontaci.nina.watch/#/submission?context=10c78596-3ea0-4867-b2fb-21fdb8e3f40c)
 
+Useful report details include:
 
+- location and institutional context;
+- supplier or vendor name;
+- visible hardware or software clues;
+- procurement documents, signage, screenshots, or public records;
+- who appears to access the data;
+- retention, oversight, and abuse risks;
+- whether the system is detection-only, identification, verification, watchlist matching, analytics, or unclear.
 
+The point is to turn deployments into inspectable evidence: claims, vendors, interfaces, procurement, sensors, data flows, and affected communities.
 
+# References dataset
 
+```text
+ REFERENCES.json ──► timeline / exhibition / research context
+ PROMPT-REFERENCES-UPDATE.txt ──► repeatable curation rules
+```
 
+[`REFERENCES.json`](REFERENCES.json) is a curated dataset of artistic, research, and activism-adjacent references around face obfuscation, adversarial appearance design, makeup-based attacks, physical-world adversarial vision, and sensor disruption.
+
+The update protocol in [`PROMPT-REFERENCES-UPDATE.txt`](PROMPT-REFERENCES-UPDATE.txt) keeps the file deduplicated, sorted, and stable. It requires canonical links, local preview-image paths, stable slugs, and a `closeness` score from `1` to `100`.
+
+Current reference links:
+
+- [Human-Imperceptible Physical Adversarial Attack for NIR Face Recognition Models](https://arxiv.org/abs/2504.15823) — Songyan Xie, Jinghang Wen, Encheng Su et al., 2025 · `research` · closeness `22`
+- [Accessorize in the Dark: A Security Analysis of Near-Infrared Face Recognition](https://doi.org/10.1007/978-3-031-51479-1_3) — Amit Cohen, Mahmood Sharif, 2024 · `research` · closeness `22`
+- [DAZZLE](https://www.michelletylicki.info/dazzle/) — Michelle Tylicki, Lauri Love, 2023 · `activism` · closeness `100`
+- [Physical-World Optical Adversarial Attacks on 3D Face Recognition](https://openaccess.thecvf.com/content/CVPR2023/html/Li_Physical-World_Optical_Adversarial_Attacks_on_3D_Face_Recognition_CVPR_2023_paper.html) — Yanjie Li, Yiquan Li, Xuelong Dai et al., 2023 · `research` · closeness `30`
+- [The Camera-Shy Hoodie](https://www.macpierce.com/the-camera-shy-hoodie) — Mac Pierce, 2023 · `artistic` · closeness `20`
+- [Shadows can be Dangerous: Stealthy and Effective Physical-world Adversarial Attack by Natural Phenomenon](https://arxiv.org/abs/2203.03818) — Yiqi Zhong, Xianming Liu, Deming Zhai et al., 2022 · `research` · closeness `35`
+- [The Dazzle Club](https://emilyroderick.com/work/the-dazzle-club/) — Evie Price, Emily Roderick, Georgina Rowlands et al., 2021 · `activism` · closeness `96`
+- [Adv-Makeup: A New Imperceptible and Transferable Attack on Face Recognition](https://arxiv.org/abs/2105.03162) — Bangjie Yin, Wenxuan Wang, Taiping Yao et al., 2021 · `research` · closeness `95`
+- [Adversarial Attacks against Face Recognition: A Comprehensive Study](https://arxiv.org/abs/2007.11709) — Fatemeh Vakhshiteh, Ahmad Nickabadi, Raghavendra Ramachandra, 2020 · `research` · closeness `82`
+- [Breaking certified defenses: Semantic adversarial examples with spoofed robustness certificates](https://arxiv.org/abs/2003.08937) — Amin Ghiasi, Ali Shafahi, Tom Goldstein, 2020 · `research` · closeness `35`
+- [VLA: A Practical Visible Light-based Attack on Face Recognition Systems in Physical World](https://doi.org/10.1145/3351261) — Meng Shen, Zelin Liao, Liehuang Zhu et al., 2019 · `research` · closeness `78`
+- [Adversarial Robustness Toolbox v1.0.0](https://arxiv.org/abs/1807.01069) — Maria-Irina Nicolae, Mathieu Sinn, Minh Ngoc Tran et al., 2018 · `research` · closeness `55`
+- [DPatch: An Adversarial Patch Attack on Object Detectors](https://arxiv.org/abs/1806.02299) — Xin Liu, Huanrui Yang, Ziwei Liu et al., 2018 · `research` · closeness `40`
+- [ShapeShifter: Robust Physical Adversarial Attack on Faster R-CNN Object Detector](https://arxiv.org/abs/1804.05810) — Shang-Tse Chen, Cory Cornelius, Jason Martin et al., 2018 · `research` · closeness `38`
+- [Adversarial Generative Nets: Neural Network Attacks on State-of-the-Art Face Recognition](https://arxiv.org/abs/1801.00349) — Mahmood Sharif, Sruti Bhagavatula, Lujo Bauer et al., 2017 · `research` · closeness `90`
+- [Adversarial Patch](https://arxiv.org/abs/1712.09665) — Tom B. Brown, Dandelion Mané, Aurko Roy et al., 2017 · `research` · closeness `50`
+- [Accessorize to a Crime: Real and Stealthy Attacks on State-of-the-Art Face Recognition](https://dl.acm.org/doi/10.1145/2976749.2978392) — Mahmood Sharif, Sruti Bhagavatula, Lujo Bauer et al., 2016 · `research` · closeness `92`
+- [HyperFace](https://adam.harvey.studio/hyperface/) — Adam Harvey, 2016 · `artistic` · closeness `92`
+- [Adversarial Manipulation of Deep Representations](https://arxiv.org/abs/1511.05122) — Sara Sabour, Yanshuai Cao, Fartash Faghri et al., 2015 · `research` · closeness `45`
+- [CV Dazzle](https://adam.harvey.studio/cvdazzle/) — Adam Harvey, 2010 · `artistic` · closeness `100`
+
+# Contributing
+
+```text
+ small patch
+ clear test
+ stable plugin API
+ documented behavior
+```
 
+Good contributions include:
 
+- new Ghostyles with clear metadata and reproducible test notes;
+- tighter unit coverage around renamed/refactored functions;
+- e2e scenarios for detection, baseline saving, overlay switching, and match-state transitions;
+- CDN self-hosting options;
+- clearer model-loading failure states;
+- accessibility and mobile UI improvements;
+- better documentation for `faceapi` vs `mediapipe` Ghostyle engines;
+- additions to [`REFERENCES.json`](REFERENCES.json) following [`PROMPT-REFERENCES-UPDATE.txt`](PROMPT-REFERENCES-UPDATE.txt).
 
-## Recent changes
-- `8296992` added the UX content for ghostyles future developer
-- `261c4f5` unified 2D and 3D ghostyle plugin
-- `0787760` thumbnail added in a list of small pictures per ID.
-- `4f47df3` infomodal added, but I'm not fully happy yet
-- `5404b34` removed some dead code
\ No newline at end of file
+Please keep claims narrow and technical: say which model, browser, lighting, camera, and threshold produced which result.
diff --git a/REFERENCES.json b/REFERENCES.json
deleted file mode 100644
index 01c0754..0000000
--- a/REFERENCES.json
+++ /dev/null
@@ -1,369 +0,0 @@
-{
-  "schema_version": "2.0",
-  "project_scope": "anti-surveillance / face-obfuscation / makeup-adjacent artistic, research, and activism references",
-  "ordering_definition": "Sort newest first. Primary key year descending; secondary key closeness descending; tertiary key title ascending.",
-  "closeness_definition": {
-    "description": "Integer 1-100 indicating how close a source is to the conceptual core of the project: face-centered anti-surveillance through appearance design, makeup, camouflage, or directly adjacent face-recognition attacks.",
-    "anchor_examples": {
-      "100": [
-        "Michelle Tylicki + Lauri Love — DAZZLE",
-        "Adam Harvey — CV Dazzle"
-      ],
-      "20": [
-        "infrared / night-vision sensor disturbance references such as The Camera-Shy Hoodie"
-      ]
-    },
-    "bands": {
-      "90_100": "direct face-centered anti-surveillance / makeup / camouflage references",
-      "70_89": "directly relevant face-recognition attack literature or strongly adjacent face-obfuscation work",
-      "40_69": "enabling or adjacent adversarial-vision methods",
-      "1_39": "peripheral or modality-shifted references, especially infrared / sensor-jamming / non-face-centric work"
-    }
-  },
-  "preview_image_definition": "Local site-root path to a preview asset intended for static visualisation. In this package the assets live under /REFERENCES-images/ as SVG files.",
-  "references": [
-    {
-      "slug": "human-imperceptible-nir-face-attack",
-      "title": "Human-Imperceptible Physical Adversarial Attack for NIR Face Recognition Models",
-      "author": [
-        "Songyan Xie",
-        "Jinghang Wen",
-        "Encheng Su",
-        "Qiucheng Yu"
-      ],
-      "year": 2025,
-      "link": "https://arxiv.org/abs/2504.15823",
-      "type": "research",
-      "closeness": 22,
-      "description": "Studio recente su perturbazioni fisiche quasi invisibili per modelli di riconoscimento facciale nel vicino infrarosso, utile come riferimento periferico sul disturbo dei sensori.",
-      "preview_image": "/REFERENCES-images/human-imperceptible-nir-face-attack.svg",
-      "visual_hint": "nir_subtle_points"
-    },
-    {
-      "slug": "accessorize-in-the-dark",
-      "title": "Accessorize in the Dark: A Security Analysis of Near-Infrared Face Recognition",
-      "author": [
-        "Amit Cohen",
-        "Mahmood Sharif"
-      ],
-      "year": 2024,
-      "link": "https://doi.org/10.1007/978-3-031-51479-1_3",
-      "type": "research",
-      "closeness": 22,
-      "description": "Analisi delle vulnerabilità dei sistemi di riconoscimento facciale nel vicino infrarosso, importante per collocare la linea di ricerca che esce dal makeup visibile e si sposta verso la sensoristica.",
-      "preview_image": "/REFERENCES-images/accessorize-in-the-dark.svg",
-      "visual_hint": "nir_points"
-    },
-    {
-      "slug": "dazzle",
-      "title": "DAZZLE",
-      "author": [
-        "Michelle Tylicki",
-        "Lauri Love"
-      ],
-      "year": 2023,
-      "link": "https://www.michelletylicki.info/dazzle/",
-      "type": "activism",
-      "closeness": 100,
-      "description": "Progetto artivista che trasforma il linguaggio del beauty in un dispositivo critico contro la sorveglianza, mettendo il volto e il gesto del trucco al centro della resistenza visiva.",
-      "preview_image": "/REFERENCES-images/dazzle.svg",
-      "visual_hint": "salon_mask"
-    },
-    {
-      "slug": "physical-world-optical-attacks-3d-face",
-      "title": "Physical-World Optical Adversarial Attacks on 3D Face Recognition",
-      "author": [
-        "Yanjie Li",
-        "Yiquan Li",
-        "Xuelong Dai",
-        "Songtao Guo",
-        "Bin Xiao"
-      ],
-      "year": 2023,
-      "link": "https://openaccess.thecvf.com/content/CVPR2023/html/Li_Physical-World_Optical_Adversarial_Attacks_on_3D_Face_Recognition_CVPR_2023_paper.html",
-      "type": "research",
-      "closeness": 30,
-      "description": "Lavoro su attacchi ottici rivolti al riconoscimento facciale 3D; amplia il campo dal piano del volto visibile a quello della profondità e della luce proiettata.",
-      "preview_image": "/REFERENCES-images/physical-world-optical-attacks-3d-face.svg",
-      "visual_hint": "projector_depth_face"
-    },
-    {
-      "slug": "camera-shy-hoodie",
-      "title": "The Camera-Shy Hoodie",
-      "author": [
-        "Mac Pierce"
-      ],
-      "year": 2023,
-      "link": "https://www.macpierce.com/the-camera-shy-hoodie",
-      "type": "artistic",
-      "closeness": 20,
-      "description": "Capo DIY che usa LED infrarossi per saturare o degradare la lettura di camere night-vision: riferimento laterale, più vicino al disturbo del sensore che al camouflage del volto.",
-      "preview_image": "/REFERENCES-images/camera-shy-hoodie.svg",
-      "visual_hint": "hoodie_ir"
-    },
-    {
-      "slug": "shadows-can-be-dangerous",
-      "title": "Shadows can be Dangerous: Stealthy and Effective Physical-world Adversarial Attack by Natural Phenomenon",
-      "author": [
-        "Yiqi Zhong",
-        "Xianming Liu",
-        "Deming Zhai",
-        "Junjun Jiang",
-        "Xiangyang Ji"
-      ],
-      "year": 2022,
-      "link": "https://arxiv.org/abs/2203.03818",
-      "type": "research",
-      "closeness": 35,
-      "description": "Introduce l’ombra come fenomeno fisico naturale capace di produrre perturbazioni credibili e difficili da notare; è una fonte laterale ma forte per pensare design e mimetismo ambientale.",
-      "preview_image": "/REFERENCES-images/shadows-can-be-dangerous.svg",
-      "visual_hint": "shadow_cast"
-    },
-    {
-      "slug": "dazzle-club",
-      "title": "The Dazzle Club",
-      "author": [
-        "Evie Price",
-        "Emily Roderick",
-        "Georgina Rowlands",
-        "Anna Hart"
-      ],
-      "year": 2021,
-      "link": "https://emilyroderick.com/work/the-dazzle-club/",
-      "type": "activism",
-      "closeness": 96,
-      "description": "Pratica collettiva tra arte, attivismo e ricerca embodied nello spazio pubblico; sposta il camouflage dal volto singolo a una coreografia sociale contro la normalizzazione della sorveglianza.",
-      "preview_image": "/REFERENCES-images/dazzle-club.svg",
-      "visual_hint": "group_walk"
-    },
-    {
-      "slug": "adv-makeup",
-      "title": "Adv-Makeup: A New Imperceptible and Transferable Attack on Face Recognition",
-      "author": [
-        "Bangjie Yin",
-        "Wenxuan Wang",
-        "Taiping Yao",
-        "Junfeng Guo",
-        "Zelun Kong",
-        "Shouhong Ding",
-        "Jilin Li",
-        "Cong Liu"
-      ],
-      "year": 2021,
-      "link": "https://arxiv.org/abs/2105.03162",
-      "type": "research",
-      "closeness": 95,
-      "description": "Paper cardine sul trucco adversarial centrato sulla regione orbitale: fondamentale per il tema makeup/riconoscimento facciale e per l’idea di blending del gesto cosmetico.",
-      "preview_image": "/REFERENCES-images/adv-makeup.svg",
-      "visual_hint": "orbital_makeup"
-    },
-    {
-      "slug": "adversarial-attacks-against-face-recognition",
-      "title": "Adversarial Attacks against Face Recognition: A Comprehensive Study",
-      "author": [
-        "Fatemeh Vakhshiteh",
-        "Ahmad Nickabadi",
-        "Raghavendra Ramachandra"
-      ],
-      "year": 2020,
-      "link": "https://arxiv.org/abs/2007.11709",
-      "type": "research",
-      "closeness": 82,
-      "description": "Survey utile per mappare tassonomie, vettori e limiti degli attacchi contro il riconoscimento facciale; funziona come vista d’insieme del territorio tecnico.",
-      "preview_image": "/REFERENCES-images/adversarial-attacks-against-face-recognition.svg",
-      "visual_hint": "survey_grid"
-    },
-    {
-      "slug": "breaking-certified-defenses-shadow-attack",
-      "title": "Breaking certified defenses: Semantic adversarial examples with spoofed robustness certificates",
-      "author": [
-        "Amin Ghiasi",
-        "Ali Shafahi",
-        "Tom Goldstein"
-      ],
-      "year": 2020,
-      "link": "https://arxiv.org/abs/2003.08937",
-      "type": "research",
-      "closeness": 35,
-      "description": "Il cosiddetto Shadow Attack lavora su perturbazioni semantiche ampie e credibili; non è face-specific, ma aiuta a leggere l’evasione come trasformazione percettiva e non solo come rumore digitale.",
-      "preview_image": "/REFERENCES-images/breaking-certified-defenses-shadow-attack.svg",
-      "visual_hint": "shadow_cast"
-    },
-    {
-      "slug": "vla-visible-light-face-attack",
-      "title": "VLA: A Practical Visible Light-based Attack on Face Recognition Systems in Physical World",
-      "author": [
-        "Meng Shen",
-        "Zelin Liao",
-        "Liehuang Zhu",
-        "Ke Xu",
-        "Xiaojiang Du"
-      ],
-      "year": 2019,
-      "link": "https://doi.org/10.1145/3351261",
-      "type": "research",
-      "closeness": 78,
-      "description": "Attacco nel mondo fisico basato su luce visibile: importante perché mostra come l’interferenza possa essere progettata come layer ottico sul volto e non solo come pattern stampato.",
-      "preview_image": "/REFERENCES-images/vla-visible-light-face-attack.svg",
-      "visual_hint": "light_beam"
-    },
-    {
-      "slug": "art-toolbox",
-      "title": "Adversarial Robustness Toolbox v1.0.0",
-      "author": [
-        "Maria-Irina Nicolae",
-        "Mathieu Sinn",
-        "Minh Ngoc Tran",
-        "Beat Buesser",
-        "Ambrish Rawat",
-        "Martin Wistuba",
-        "Valentina Zantedeschi",
-        "Nathalie Baracaldo",
-        "Bryant Chen",
-        "Heiko Ludwig",
-        "Ian M. Molloy",
-        "Ben Edwards"
-      ],
-      "year": 2018,
-      "link": "https://arxiv.org/abs/1807.01069",
-      "type": "research",
-      "closeness": 55,
-      "description": "Framework di riferimento per attacchi e difese adversariali; non è una fonte estetica, ma è la cornice tecnica che connette molti dei paper operativi del dataset.",
-      "preview_image": "/REFERENCES-images/art-toolbox.svg",
-      "visual_hint": "toolbox"
-    },
-    {
-      "slug": "dpatch",
-      "title": "DPatch: An Adversarial Patch Attack on Object Detectors",
-      "author": [
-        "Xin Liu",
-        "Huanrui Yang",
-        "Ziwei Liu",
-        "Linghao Song",
-        "Hai Li",
-        "Yiran Chen"
-      ],
-      "year": 2018,
-      "link": "https://arxiv.org/abs/1806.02299",
-      "type": "research",
-      "closeness": 40,
-      "description": "Estende la logica della patch agli object detector, mostrando come una piccola superficie progettata possa sabotare localizzazione e classificazione.",
-      "preview_image": "/REFERENCES-images/dpatch.svg",
-      "visual_hint": "detector_box_patch"
-    },
-    {
-      "slug": "shapeshifter",
-      "title": "ShapeShifter: Robust Physical Adversarial Attack on Faster R-CNN Object Detector",
-      "author": [
-        "Shang-Tse Chen",
-        "Cory Cornelius",
-        "Jason Martin",
-        "Duen Horng Chau"
-      ],
-      "year": 2018,
-      "link": "https://arxiv.org/abs/1804.05810",
-      "type": "research",
-      "closeness": 38,
-      "description": "Riferimento classico sugli attacchi fisici robusti ai detector: meno vicino al volto, ma utile per la genealogia delle perturbazioni materiali e stampabili.",
-      "preview_image": "/REFERENCES-images/shapeshifter.svg",
-      "visual_hint": "object_shape_distort"
-    },
-    {
-      "slug": "agns-face-recognition",
-      "title": "Adversarial Generative Nets: Neural Network Attacks on State-of-the-Art Face Recognition",
-      "author": [
-        "Mahmood Sharif",
-        "Sruti Bhagavatula",
-        "Lujo Bauer",
-        "Michael K. Reiter"
-      ],
-      "year": 2017,
-      "link": "https://arxiv.org/abs/1801.00349",
-      "type": "research",
-      "closeness": 90,
-      "description": "Estende la genealogia degli occhiali adversarial con una generazione più robusta e scalabile, facendo degli accessori facciali un vero medium di attacco fisico.",
-      "preview_image": "/REFERENCES-images/agns-face-recognition.svg",
-      "visual_hint": "generated_glasses"
-    },
-    {
-      "slug": "adversarial-patch",
-      "title": "Adversarial Patch",
-      "author": [
-        "Tom B. Brown",
-        "Dandelion Mané",
-        "Aurko Roy",
-        "Martín Abadi",
-        "Justin Gilmer"
-      ],
-      "year": 2017,
-      "link": "https://arxiv.org/abs/1712.09665",
-      "type": "research",
-      "closeness": 50,
-      "description": "Paper seminale sulla patch universale e stampabile; è meno face-centered ma fondamentale per capire la logica dell’oggetto grafico che devia l’attenzione algoritmica.",
-      "preview_image": "/REFERENCES-images/adversarial-patch.svg",
-      "visual_hint": "patch_sticker"
-    },
-    {
-      "slug": "accessorize-to-a-crime",
-      "title": "Accessorize to a Crime: Real and Stealthy Attacks on State-of-the-Art Face Recognition",
-      "author": [
-        "Mahmood Sharif",
-        "Sruti Bhagavatula",
-        "Lujo Bauer",
-        "Michael K. Reiter"
-      ],
-      "year": 2016,
-      "link": "https://dl.acm.org/doi/10.1145/2976749.2978392",
-      "type": "research",
-      "closeness": 92,
-      "description": "Punto di svolta sugli occhiali adversarial fisici: dimostra che un accessorio quotidiano può diventare una superficie di attacco mirata contro la biometria facciale.",
-      "preview_image": "/REFERENCES-images/accessorize-to-a-crime.svg",
-      "visual_hint": "glasses"
-    },
-    {
-      "slug": "hyperface",
-      "title": "HyperFace",
-      "author": [
-        "Adam Harvey"
-      ],
-      "year": 2016,
-      "link": "https://adam.harvey.studio/hyperface/",
-      "type": "artistic",
-      "closeness": 92,
-      "description": "Tessuto/camouflage basato su false facce che abbassa l’affidabilità della detection; riferimento artistico cruciale per pensare il pattern come rumore semantico.",
-      "preview_image": "/REFERENCES-images/hyperface.svg",
-      "visual_hint": "false_faces"
-    },
-    {
-      "slug": "adversarial-manipulation-deep-representations",
-      "title": "Adversarial Manipulation of Deep Representations",
-      "author": [
-        "Sara Sabour",
-        "Yanshuai Cao",
-        "Fartash Faghri",
-        "David J. Fleet"
-      ],
-      "year": 2015,
-      "link": "https://arxiv.org/abs/1511.05122",
-      "type": "research",
-      "closeness": 45,
-      "description": "Fonte teorica dietro le feature adversaries: importante per capire come si possa spingere un’immagine verso le rappresentazioni interne di un’altra senza copiarne l’aspetto.",
-      "preview_image": "/REFERENCES-images/adversarial-manipulation-deep-representations.svg",
-      "visual_hint": "latent_match"
-    },
-    {
-      "slug": "cv-dazzle",
-      "title": "CV Dazzle",
-      "author": [
-        "Adam Harvey"
-      ],
-      "year": 2010,
-      "link": "https://adam.harvey.studio/cvdazzle/",
-      "type": "artistic",
-      "closeness": 100,
-      "description": "Progetto fondativo del camouflage per computer vision: makeup, hairstyling e asimmetria geometrica usati per interrompere i segnali che i detector facciali cercano nel volto umano.",
-      "preview_image": "/REFERENCES-images/cv-dazzle.svg",
-      "visual_hint": "dazzle_face"
-    }
-  ]
-}
diff --git a/about.html b/about.html
new file mode 100644
index 0000000..beb0f36
--- /dev/null
+++ b/about.html
@@ -0,0 +1,240 @@
+<!doctype html>
+<html lang="en">
+
+<head>
+  <meta charset="utf-8" />
+  <meta name="viewport" content="width=device-width, initial-scale=1" />
+  <title>Vision & About — Ghostmaxxing</title>
+  <meta name="description"
+    content="Ghostmaxxing is a browser-based public research platform for testing face-recognition camouflage, running workshops, and investigating real-world biometric surveillance." />
+  <link rel="stylesheet" href="styles/pages.css" />
+  <link rel="stylesheet" href="styles/content-pages.css" />
+</head>
+
+<body>
+  <main class="editorial-homepage content-page editorial-homepage--svg-footer-band">
+    <div class="editorial-homepage__shell">
+      <header class="editorial-homepage__header">
+        <a class="wordmark" href="index.html" aria-label="Ghostmaxxing homepage">
+          <span class="wordmark__title">Ghostmaxxing</span>
+          <span class="wordmark__subtitle">
+            Privacy activism meets public-research platform
+          </span>
+        </a>
+
+        <nav class="editorial-homepage__nav" aria-label="Project navigation">
+          <a class="navlink" href="https://github.com/vecna/ghostati" target="_blank" rel="noopener noreferrer">Code</a>
+          <a class="navlink" href="about.html" aria-current="page">Vision &amp; About</a>
+          <a class="cta-inline" href="report.html">How to <i>leak</i> to us<span aria-hidden="true">↗</span></a>
+        </nav>
+      </header>
+
+      <section class="content-page__header-area" aria-labelledby="page-title">
+        <span class="content-page__kicker">Vision &amp; About</span>
+        <h1 id="page-title">Make the machine-readable face <i>contestable</i>.</h1>
+        <p class="content-page__lead">
+          Ghostmaxxing is a browser-based workshop assistant and public research platform for testing
+          face-recognition camouflage.
+        </p>
+        <p class="content-page__deck">
+          It grows out of privacy activism, adversarial makeup workshops, and the need to investigate which biometric
+          systems are actually used in public space.
+        </p>
+      </section>
+
+      <div class="content-page__layout">
+        <article class="content-page__main">
+          <section class="content-section" id="workshop-assistant">
+            <h2>Workshop assistant.</h2>
+            <p>
+              Designed to support group workshops, where participants play with makeup and learn how different faces
+              read to a recognition model: point a browser camera at your face, record a baseline, apply a Ghostyle,
+              and compare how the local recognition pipeline reacts.
+            </p>
+            <p>
+              The goal is to lower the barrier to entry. You don't need a makeup specialist to take part &mdash; only
+              water-based colors, safe enough for kids. The system runs locally: it uses the webcam or a phone
+              camera, keeps everything on-device, and sends nothing anywhere else, except what you choose to share
+              with us.
+            </p>
+            <p>
+              The project is built to run in a phone browser, because makeup is usually applied in front of a mirror,
+              not at a desktop workstation.
+            </p>
+            <div class="callout">
+              <p>
+                Two kinds of success exist: the face stops being detected as a face at all, or the face is still
+                detected but its extracted points differ enough from the clean-face baseline that matching fails.
+              </p>
+            </div>
+            <h3>What can be sent to us?</h3>
+            <p>Three things mostly:</p>
+            <ol>
+              <li>
+                Before/after images of a successful attempt, sent via the
+                <!-- TODO: link/insert the share icon from lab.html once it's finalized -->
+                share icon in the lab.
+              </li>
+              <li>A one-second video of a successful attempt.</li>
+              <li>
+                Ghostyles: modular camouflage plugins that others can study, adapt, or retest.
+                <!-- TODO: write the Ghostyle plugin docs and link them here (e.g. /docs) -->
+                Build your own by following the Ghostyle documentation.
+              </li>
+            </ol>
+          </section>
+
+          <section class="content-section" id="before-after">
+            <h2>From before/after experiments to Ghostyles.</h2>
+            <p>
+              This part matters: we want a system that can run a "diff" between an after image and a before image,
+              and turn the result into a Ghostyle plugin that others can upload and retest.
+            </p>
+            <div class="callout">
+              <p>
+                These approaches are experimental. We can't guarantee they will work &mdash; and a human reviewing
+                camera footage can often still recognize someone even when they're wearing makeup. A browser result
+                is a local, conditional finding, not a general protection claim.
+              </p>
+            </div>
+          </section>
+
+          <section class="content-section" id="origin">
+            <h2>Where it comes from.</h2>
+            <p>
+              Ghostmaxxing is the international evolution of an experiment run by NINA.watch. As part of the
+              Universal Digital Union, <a href="https://sindacato.nina.watch/it/iniziative/ghostati/">Ghòstati</a>
+              (<i>become a ghost</i>) is a workshop we keep repeating to explore adversarial makeup as a way to
+              resist facial recognition. The project began in May 2026 during the NINA festival, in
+              <a href="https://sindacato.nina.watch/it/blog/ghostati-workshop-maggio-2026/">Milan and Rome</a>. We
+              learned from <a href="https://www.michelletylicki.info/">Michelle Tylicki</a>, an artist who has run
+              this kind of workshop before us &mdash; we added the application development, and now this wider
+              vision.
+            </p>
+            <p class="callout">
+              Tylicki's <a href="https://www.michelletylicki.info/dazzle/">DAZZLE</a>, developed with Lauri Love, is
+              an art-and-tech installation and interactive tool for teaching Computer Vision Dazzle, also known as
+              anti-surveillance makeup. Ghostmaxxing inherits that workshop energy, but turns it into a browser-based
+              system people can run on their own devices.
+            </p>
+          </section>
+
+          <section class="content-section" id="vision">
+            <h2>Vision.</h2>
+            <ul>
+              <li>Make the workshop easier to run.</li>
+              <li>Make adversarial makeup a popular practice, not a niche one.</li>
+              <li>Collect evidence of what actually works.</li>
+              <li>Support new forms &mdash; clothing, 3D-printed objects, and beyond.</li>
+              <li>Support new research.</li>
+              <li>Use network effects to test everywhere, share results, and keep improving.</li>
+            </ul>
+          </section>
+
+          <section class="content-section" id="research-question">
+            <h2>The research question is minimal camouflage.</h2>
+            <p>
+              The core question is simple: how little makeup is needed to disrupt a face-recognition pipeline?
+            </p>
+            <p>
+              Less makeup matters because it lowers the barrier to practice. It makes the technique easier to learn,
+              easier to repeat, and easier to move from exceptional workshop performance toward everyday culture. The
+              hope is not to sell invisibility; it is to make adversarial makeup legible enough to become pop practice.
+            </p>
+            <h3>Success should be shareable.</h3>
+            <p>
+              Ghostmaxxing includes sharing-oriented functions because a successful look is also a teachable pattern.
+              People who want to share a one-second video, a before/after comparison, or a Ghostyle result can help
+              others understand what worked and what still needs to be tested.
+            </p>
+          </section>
+
+          <section class="content-section" id="leaks">
+            <h2>Why leaking matters.</h2>
+            <p>
+              The lab can test browser pipelines, but real-world facial recognition is a supply chain: camera
+              hardware, edge devices, model vendors, watchlists, matching systems, dashboards, alerts, metadata,
+              operators, procurement contracts, and retention rules.
+            </p>
+            <p>
+              <a href="report.html">Leak to us</a> exists to understand that hidden chain. Which technologies are
+              deployed? Who sells them? Who maintains them? What metadata is produced between capture and decision?
+              Where does it go? Who can access it? Tell us what you know, only if it is safe for you to do so &mdash;
+              and please don't include unnecessary personal data.
+            </p>
+            <p>
+              Some allies may be forced to work on or near these systems. If they can safely share information, they
+              may help us understand whether Ghostyles and adversarial makeup are actually affecting real
+              deployments, or whether the resistance needs to change.
+            </p>
+            <div class="callout">
+              <p>
+                In Europe, real-time remote biometric identification in publicly accessible spaces for law
+                enforcement is treated as a prohibited AI practice, subject to narrow exceptions and safeguards. That
+                legal frame still leaves a practical question: what is actually being deployed, and under whose
+                control?
+              </p>
+            </div>
+            <h3>A ban that keeps getting reopened.</h3>
+            <p>
+              The <a href="https://reclaimyourface.eu">Reclaim Your Face</a> campaign has spent years pushing for a
+              ban on biometric mass surveillance, and it helped get real-time remote biometric identification treated
+              as prohibited in EU law. But the fight didn't end with the text of the regulation: national security
+              agendas keep reopening the exceptions, arguing for carve-outs around major events, "special" cases, and
+              pilot deployments. <a
+                href="https://reclaimyourface.eu/biometric-surveillance-in-the-czech-republic-the-ministry-of-the-interior-is-trying-to-circumvent-the-artificial-intelligence-act/">Reporting
+                from the campaign</a>
+              has documented national authorities attempting to work around the AI Act's limits rather than comply
+              with them. A ban that keeps getting quietly worked around by fear-mongering, securitarian framing needs
+              the same kind of public pressure and evidence-gathering that got it written in the first place &mdash;
+              which is part of why the reporting node exists here too.
+            </p>
+          </section>
+
+          <section class="content-section" id="resources">
+            <h2>Read, test, report.</h2>
+            <p>
+              Ghostmaxxing is strongest when the lab, the archive, and the reporting channel work together. Test
+              techniques locally, read the lineage of anti-biometric appearance design, and help document the real
+              infrastructure when it appears in the world.
+            </p>
+            <div class="resource-grid">
+              <a class="resource-card" href="ghostati.html">
+                <h3>Open the lab</h3>
+                <p>Run local browser tests, try Ghostyles, and compare before/after behaviour.</p>
+              </a>
+              <a class="resource-card" href="/references/">
+                <h3>Research references</h3>
+                <p>Trace the artistic, technical, and activist lineage of face-obfuscation work.</p>
+              </a>
+              <a class="resource-card" href="/docs/">
+                <h3>Technical documentation</h3>
+                <p>Inspect modules, plugin logic, and implementation details.</p>
+              </a>
+              <a class="resource-card" href="report.html">
+                <h3>Leak to us</h3>
+                <p>Help investigate the real-world pipeline and supply chain of face recognition.</p>
+              </a>
+            </div>
+          </section>
+        </article>
+
+        <aside class="content-page__aside" aria-label="Page sections">
+          <h2>On this page</h2>
+          <nav>
+            <a href="#workshop-assistant">Workshop assistant</a>
+            <a href="#before-after">Before / after</a>
+            <a href="#origin">Origin</a>
+            <a href="#research-question">Research question</a>
+            <a href="#leaks">Why leaking matters</a>
+            <a href="#resources">Resources</a>
+          </nav>
+        </aside>
+      </div>
+    </div>
+
+    <div class="camera-band" aria-hidden="true"></div>
+  </main>
+</body>
+
+</html>
\ No newline at end of file
diff --git a/ghostati-docs.html b/ghostati-docs.html
deleted file mode 100644
index e680265..0000000
--- a/ghostati-docs.html
+++ /dev/null
@@ -1,194 +0,0 @@
-<!doctype html>
-<html lang="it">
-
-<head>
-  <meta charset="utf-8" />
-  <meta name="viewport" content="width=device-width, initial-scale=1" />
-  <title>Ghòstati - Guida per Sviluppatori</title>
-  <link rel="stylesheet" href="styles/ghostati-docs.css" />
-  <link rel="stylesheet" href="styles/index.css" />
-
-  <link rel="icon" type="image/x-icon" href="favicon.ico" />
-  <link rel="sitemap" type="application/xml" title="Sitemap" href="sitemap.xml" />
-</head>
-
-<body>
-  <div class="page">
-
-    <header class="ghost-topbar" id="ghostTopbar">
-
-      <button class="ghost-topbar__toggle" id="ghostMenuToggle" type="button" aria-expanded="false"
-        aria-controls="ghostTopbarMenu" aria-label="Apri menu">
-        <span></span>
-        <span></span>
-        <span></span>
-      </button>
-
-      <nav class="ghost-topbar__nav" aria-label="Workshop e progetto">
-        <ul class="ghost-topbar__links" id="ghostTopbarMenu">
-          <li><a href="ghostati.html">Ghòstati!</a></li>
-          <li><a href="https://github.com/vecna/ghostati" target="_blank" rel="noopener noreferrer">codice</a></li>
-          <li><a href="https://sindacato.nina.watch/">progetto</a></li>
-        </ul>
-      </nav>
-
-    </header>
-
-    <h1>Documentazione</h1>
-    <div class="subtitle">Sviluppa il tuo Ghostyle</div>
-
-
-    <section>
-      <h2>Architettura ad Alto Livello</h2>
-      <p>Il sistema "Ghòstati" utilizza un'<strong>architettura a plugin modulari</strong> pensata per separare il
-        nucleo di tracciamento facciale dagli effetti visivi (i "Ghostyles").</p>
-
-      <p>Tecnicamente, l'engine principale è una web app Vanilla JS che inizializza la webcam, e utilizza
-        <code>face-api.js</code> per inferire in tempo reale i <em>Landmarks a 68 punti</em> del viso. Il loop di
-        rendering è gestito autonomamente per garantire performance scalabili (FPS controllati).
-      </p>
-
-      <p>Og intanto, l'engine richiama l'effetto attivo (il plugin). Ogni <strong>Ghostyle</strong> è un ES Module
-        (<code>.js</code>) caricato dinamicamente via <code>import(url)</code>. Questo significa che i Ghostyle possono
-        essere creati come file locali indipendenti, o addirittura hostati su server remoti e passati via URL.</p>
-
-      <p>L'engine fornisce al plugin due cose: le coordinate del volto aggiornate e le API globali offerte da
-        <code>window.Ghostati</code> per semplificare drasticamente la geometria e il disegno su un Canvas dedicato
-        (livello di overlay).
-      </p>
-    </section>
-
-    <section>
-      <h2>Template Didattico</h2>
-      <p>Un plugin Ghòstyle per funzionare deve unicamente esportare un paio di funzioni specifiche (gli "Event Hooks").
-        Questo snippet di codice rappresenta lo scheletro di base per cominciare a sviluppare.</p>
-
-      <pre><code>// @name Effetto Base
-
-/**
- * [BLOCCO INIT]
- * Funzione di inizializzazione locale (Opzionale).
- * Chiamata *una sola volta* quando il Ghostyle viene importato in memoria.
- */
-export function onInit() {
-  Ghostati.log('Effetto base caricato correttamente!', 'BASE_STYLE');
-}
-
-/**
- * [BLOCCO DRAW]
- * Funzione principale di rendering.
- * L'Engine richiama questa funzione iterativamente nel RAF loop.
- * 
- * @param {CanvasRenderingContext2D} ctx - Il contesto 2D nativo.
- * @param {faceapi.FaceLandmarks68} landmarks - Collezione dei landmark spaziali del viso.
- * @param {faceapi.Box} box - Il Bounding Box che riquadra tutto il viso.
- */
-export function onDraw(ctx, landmarks, box) {
-  // 1. Estrazione Segmenti (Array di {x,y})
-  const leftEye = landmarks.getLeftEye();
-  const rightEye = landmarks.getRightEye();
-
-  // 2. Logging temporaneo (opzionale) o manipolazione
-  // Attenzione: usare Ghostati.log qui scriverà messaggi ad alto rate / frame.
-  
-  // 3. Staging via ctx standard
-  ctx.fillStyle = 'rgba(255, 122, 122, 0.45)';
-  ctx.strokeStyle = 'rgba(255, 122, 122, 0.9)';
-
-  // 4. Rendering tramite libreria Helper
-  Ghostati.drawClosedPath(ctx, leftEye, ctx.fillStyle, ctx.strokeStyle, 2);
-  Ghostati.drawClosedPath(ctx, rightEye, ctx.fillStyle, ctx.strokeStyle, 2);
-}
-
-/**
- * [BLOCCO CLEAR]
- * Funzione di smontaggio / teardown (Opzionale).
- * L'Engine svuota automaticamente il Canvas. Usalo se hai timer o memorie da annullare.
- */
-export function onClear(ctx) {
-  console.log("Ripristino pulizia plugin base.");
-}
-</code></pre>
-    </section>
-
-    <section>
-      <h2>Spiegazione dei Blocchi</h2>
-
-      <div class="block-explain">
-        <strong>1. Metadati Commentati</strong>
-        La prima riga <code>// @name [Nome]</code> è letta dall'Engine in fase di fetch come fallback se il plugin non
-        ha un id chiaro o necessita di una label UI amichevole. Non è strettamente javascript-eseguibile, ma va in cima
-        al modulo!
-      </div>
-
-      <div class="block-explain">
-        <strong>2. Blocco INIT (<code>export function onInit</code>)</strong>
-        È l'Entry Point primario eseguito all'iniezione nel DOM/Browser dal Plugin Manager. Idealmente usato per
-        scaricare asset aggiuntivi o notificare l'interfaccia via <code>Ghostati.log</code>. Può essere asincrono
-        (<code>async/await</code>).
-      </div>
-
-      <div class="block-explain">
-        <strong>3. Blocco DRAW (<code>export function onDraw</code>)</strong>
-        È il cuore pulsante del tuo scrip AR. L'argomento <code>landmarks</code> espone i metodi standard di face-api
-        per ottenere array di coordinate <code>{x,y}</code> suddivise semanticamente: <code>getLeftEye()</code>,
-        <code>getJawOutline()</code>, <code>getNose()</code>, <code>getMouth()</code>. È tutto disegnato su un livello
-        trasparente perfettamente riallineato sulla webcam.
-      </div>
-
-      <div class="block-explain">
-        <strong>4. Utilizzo Metodi <code>Ghostati</code></strong>
-        Piuttosto che calcolare le curve Bezier punto per punto, usa le Utility messe a disposizione nell'interfaccia
-        <code>window.Ghostati</code> come <code>drawClosedPath</code>. Ci sono anche funzioni di morphing sofisticate
-        come <code>Ghostati.expandEyePolygon</code> o <code>Ghostati.drawEyeWing</code> per riprodurre eyeliner e sagome
-        senza sforzo matematico.
-      </div>
-
-      <div class="block-explain">
-        <strong>5. Blocco CLEAR (<code>export function onClear</code>)</strong>
-        L'Engine passa a questo Hook prima di effettuare uno "switch" verso un altro plugin, garantendovi un life-cycle
-        appropriato in cui eliminare custom listeners o fermare calcoli costosi derivati se necessario.
-      </div>
-    </section>
-
-    <section>
-      <h2>Modalità Diagnostica (Test CV Dazzle)</h2>
-      <p>Ghòstati non si limita ad applicare il trucco AR in modo cosmetico, ma include uno strumento di
-        <strong>Diagnostica e Analisi delle Vulnerabilità</strong> progettato per valutare la reale efficacia del
-        <em>Ghostyle</em> nel neutralizzare gli algoritmi di Facial Recognition (CV Dazzle).
-      </p>
-
-      <div class="block-explain">
-        <strong>Pulsante "Scansiona Trucco" e Compositing Off-screen</strong>
-        Quando un Ghostyle è attivo, il normale pulsante maschera la sua funzione primaria e si trasforma in "Scansiona
-        Trucco". Premendolo, il sistema:
-        <ol style="margin-top: 8px; margin-bottom: 8px; padding-left: 20px;">
-          <li>Genera un <strong>canvas off-screen</strong> invisibile all'utente.</li>
-          <li>Esegue un <em>compositing</em> fondendo pixel per pixel il feed reale della webcam con l'output grafico
-            del trucco (mantenendone inalterate le opacità e i colori originali).</li>
-          <li>Forza il motore neurale (<code>face-api</code>) a scansionare questa nuova immagine sintetica per vedere
-            se riesce ad estrarre la stessa identità.</li>
-        </ol>
-        Durante questo processo, l'overlay a schermo rimane "bloccato" per continuare a fungere da specchio di guida,
-        senza scomodi fade-out visivi.
-      </div>
-
-      <div class="block-explain">
-        <strong>Opacità e Colori: Il fattore di difficoltà</strong>
-        Ai fini del test, il Canvas composito rispetta testualmente gli stili <code>rgba()</code> del tuo codice (non li
-        forza al 100% di opacità). Questo significa che se usi colori troppo tenui, c'è un'alta probabilità che il
-        sistema biometrico riesca a "vedere attraverso" il tratto. Nell'archivio locale a schermo è annotata in tempo
-        reale "l'Opacità stimata" del trucco.
-      </div>
-
-      <div class="block-explain">
-        <strong>Pulsante "Copia Volto Truccato"</strong>
-        Una volta che la simulazione è conclusa con successo, un pulsante speciale si accende. Cliccandolo, l'immagina
-        composita generata <em>(volto della webcam alterato permanentemente dal Ghostyle sovrapposto)</em> viene salvata
-        comodamente negli Appunti del sistema operativo, pronta per essere ispezionata.
-      </div>
-    </section>
-  </div>
-</body>
-
-</html>
\ No newline at end of file
diff --git a/ghostyles/00-template.js b/ghostyles/00-template.js
index d4953ec..e24584c 100644
--- a/ghostyles/00-template.js
+++ b/ghostyles/00-template.js
@@ -4,7 +4,7 @@
  * @version      2.0.0
  * @author       NINA
  * @release_date 2026-06-29
- * @description  Template canonico per plugin Ghòstati: esempio minimo con callback 2D + UV.
+ * @description  Template canonico per plugin Ghostmaxxing: esempio minimo con callback 2D + UV.
  * ==/Ghostyle==
  */
 
diff --git a/images/ghostati-hero.png b/images/ghostati-hero.png
deleted file mode 100644
index 24e738c..0000000
Binary files a/images/ghostati-hero.png and /dev/null differ
diff --git a/images/ghostmaxxing-homepage.webp b/images/ghostmaxxing-homepage.webp
new file mode 100644
index 0000000..487e955
Binary files /dev/null and b/images/ghostmaxxing-homepage.webp differ
diff --git a/images/homepage-camera-band.svg b/images/homepage-camera-band.svg
new file mode 100644
index 0000000..b1f1a38
--- /dev/null
+++ b/images/homepage-camera-band.svg
@@ -0,0 +1,502 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 2800 420" width="2800" height="420" role="img" aria-labelledby="title desc">
+<title id="title">Camera flowers footer band</title>
+<desc id="desc">A wide decorative band of surveillance cameras growing from stems and leaves, designed as a bottom-aligned footer motif.</desc>
+<rect width="2800" height="420" fill="none"/>
+
+
+<g transform="translate(-80 420)">
+  <path d="M0 0 C34 -75 58 -64 75 -94 C53 -21 22 -8 0 0 Z" fill="#063f32"/>
+  <path d="M20 -8 C-34 -52 -51 -52 -68 -78 C-48 -15 -15 -6 20 -8 Z" fill="#063f32" opacity="0.92"/>
+</g>
+
+<g transform="translate(5 420)">
+  <path d="M0 0 C45 -112 76 -95 99 -140 C70 -31 29 -11 0 0 Z" fill="#0b4b3a"/>
+  <path d="M20 -8 C-45 -78 -68 -78 -90 -116 C-63 -22 -20 -9 20 -8 Z" fill="#0b4b3a" opacity="0.92"/>
+</g>
+
+<g transform="translate(90 420)">
+  <path d="M0 0 C56 -149 95 -127 123 -186 C87 -42 36 -15 0 0 Z" fill="#155b44"/>
+  <path d="M20 -8 C-56 -104 -84 -104 -112 -155 C-78 -30 -25 -12 20 -8 Z" fill="#155b44" opacity="0.92"/>
+</g>
+
+<g transform="translate(175 420)">
+  <path d="M0 0 C37 -186 63 -158 81 -232 C57 -52 24 -19 0 0 Z" fill="#092f2a"/>
+  <path d="M20 -8 C-37 -130 -56 -130 -74 -193 C-52 -37 -17 -15 20 -8 Z" fill="#092f2a" opacity="0.92"/>
+</g>
+
+<g transform="translate(260 420)">
+  <path d="M0 0 C48 -88 82 -75 106 -110 C74 -25 31 -9 0 0 Z" fill="#063f32"/>
+  <path d="M20 -8 C-48 -62 -72 -62 -96 -92 C-67 -18 -22 -7 20 -8 Z" fill="#063f32" opacity="0.92"/>
+</g>
+
+<g transform="translate(345 420)">
+  <path d="M0 0 C59 -125 100 -106 130 -156 C91 -35 38 -12 0 0 Z" fill="#0b4b3a"/>
+  <path d="M20 -8 C-59 -88 -88 -88 -118 -130 C-83 -25 -27 -10 20 -8 Z" fill="#0b4b3a" opacity="0.92"/>
+</g>
+
+<g transform="translate(430 420)">
+  <path d="M0 0 C40 -162 68 -138 88 -202 C62 -45 26 -16 0 0 Z" fill="#155b44"/>
+  <path d="M20 -8 C-40 -113 -60 -113 -80 -168 C-56 -32 -18 -13 20 -8 Z" fill="#155b44" opacity="0.92"/>
+</g>
+
+<g transform="translate(515 420)">
+  <path d="M0 0 C51 -199 87 -169 112 -249 C79 -56 33 -20 0 0 Z" fill="#092f2a"/>
+  <path d="M20 -8 C-51 -139 -76 -139 -102 -207 C-71 -40 -23 -16 20 -8 Z" fill="#092f2a" opacity="0.92"/>
+</g>
+
+<g transform="translate(600 420)">
+  <path d="M0 0 C62 -101 105 -86 136 -126 C96 -28 40 -10 0 0 Z" fill="#063f32"/>
+  <path d="M20 -8 C-62 -71 -93 -71 -124 -105 C-87 -20 -28 -8 20 -8 Z" fill="#063f32" opacity="0.92"/>
+</g>
+
+<g transform="translate(685 420)">
+  <path d="M0 0 C43 -138 73 -117 95 -172 C67 -39 28 -14 0 0 Z" fill="#0b4b3a"/>
+  <path d="M20 -8 C-43 -97 -64 -97 -86 -144 C-60 -28 -19 -11 20 -8 Z" fill="#0b4b3a" opacity="0.92"/>
+</g>
+
+<g transform="translate(770 420)">
+  <path d="M0 0 C54 -175 92 -149 119 -219 C84 -49 35 -18 0 0 Z" fill="#155b44"/>
+  <path d="M20 -8 C-54 -122 -81 -122 -108 -182 C-76 -35 -24 -14 20 -8 Z" fill="#155b44" opacity="0.92"/>
+</g>
+
+<g transform="translate(855 420)">
+  <path d="M0 0 C35 -77 60 -65 77 -96 C54 -22 23 -8 0 0 Z" fill="#092f2a"/>
+  <path d="M20 -8 C-35 -54 -52 -54 -70 -80 C-49 -15 -16 -6 20 -8 Z" fill="#092f2a" opacity="0.92"/>
+</g>
+
+<g transform="translate(940 420)">
+  <path d="M0 0 C46 -114 78 -97 101 -142 C71 -32 30 -11 0 0 Z" fill="#063f32"/>
+  <path d="M20 -8 C-46 -80 -69 -80 -92 -119 C-64 -23 -21 -9 20 -8 Z" fill="#063f32" opacity="0.92"/>
+</g>
+
+<g transform="translate(1025 420)">
+  <path d="M0 0 C57 -151 97 -128 125 -189 C88 -42 37 -15 0 0 Z" fill="#0b4b3a"/>
+  <path d="M20 -8 C-57 -106 -86 -106 -114 -157 C-80 -30 -26 -12 20 -8 Z" fill="#0b4b3a" opacity="0.92"/>
+</g>
+
+<g transform="translate(1110 420)">
+  <path d="M0 0 C38 -188 65 -160 84 -235 C59 -53 25 -19 0 0 Z" fill="#155b44"/>
+  <path d="M20 -8 C-38 -132 -57 -132 -76 -196 C-53 -38 -17 -15 20 -8 Z" fill="#155b44" opacity="0.92"/>
+</g>
+
+<g transform="translate(1195 420)">
+  <path d="M0 0 C49 -90 83 -76 108 -112 C76 -25 32 -9 0 0 Z" fill="#092f2a"/>
+  <path d="M20 -8 C-49 -63 -74 -63 -98 -94 C-69 -18 -22 -7 20 -8 Z" fill="#092f2a" opacity="0.92"/>
+</g>
+
+<g transform="translate(1280 420)">
+  <path d="M0 0 C60 -127 102 -108 132 -159 C93 -36 39 -13 0 0 Z" fill="#063f32"/>
+  <path d="M20 -8 C-60 -89 -90 -89 -120 -132 C-84 -25 -27 -10 20 -8 Z" fill="#063f32" opacity="0.92"/>
+</g>
+
+<g transform="translate(1365 420)">
+  <path d="M0 0 C41 -164 70 -139 90 -205 C64 -46 27 -16 0 0 Z" fill="#0b4b3a"/>
+  <path d="M20 -8 C-41 -115 -62 -115 -82 -171 C-57 -33 -18 -13 20 -8 Z" fill="#0b4b3a" opacity="0.92"/>
+</g>
+
+<g transform="translate(1450 420)">
+  <path d="M0 0 C52 -201 88 -171 114 -251 C81 -56 34 -20 0 0 Z" fill="#155b44"/>
+  <path d="M20 -8 C-52 -141 -78 -141 -104 -209 C-73 -40 -23 -16 20 -8 Z" fill="#155b44" opacity="0.92"/>
+</g>
+
+<g transform="translate(1535 420)">
+  <path d="M0 0 C63 -103 107 -88 139 -129 C98 -29 41 -10 0 0 Z" fill="#092f2a"/>
+  <path d="M20 -8 C-63 -72 -94 -72 -126 -107 C-88 -21 -28 -8 20 -8 Z" fill="#092f2a" opacity="0.92"/>
+</g>
+
+<g transform="translate(1620 420)">
+  <path d="M0 0 C44 -140 75 -119 97 -175 C68 -39 29 -14 0 0 Z" fill="#063f32"/>
+  <path d="M20 -8 C-44 -98 -66 -98 -88 -146 C-62 -28 -20 -11 20 -8 Z" fill="#063f32" opacity="0.92"/>
+</g>
+
+<g transform="translate(1705 420)">
+  <path d="M0 0 C55 -177 94 -150 121 -221 C85 -50 36 -18 0 0 Z" fill="#0b4b3a"/>
+  <path d="M20 -8 C-55 -124 -82 -124 -110 -184 C-77 -35 -25 -14 20 -8 Z" fill="#0b4b3a" opacity="0.92"/>
+</g>
+
+<g transform="translate(1790 420)">
+  <path d="M0 0 C36 -79 61 -67 79 -99 C56 -22 23 -8 0 0 Z" fill="#155b44"/>
+  <path d="M20 -8 C-36 -55 -54 -55 -72 -82 C-50 -16 -16 -6 20 -8 Z" fill="#155b44" opacity="0.92"/>
+</g>
+
+<g transform="translate(1875 420)">
+  <path d="M0 0 C47 -116 80 -99 103 -145 C73 -32 31 -12 0 0 Z" fill="#092f2a"/>
+  <path d="M20 -8 C-47 -81 -70 -81 -94 -121 C-66 -23 -21 -9 20 -8 Z" fill="#092f2a" opacity="0.92"/>
+</g>
+
+<g transform="translate(1960 420)">
+  <path d="M0 0 C58 -153 99 -130 128 -191 C90 -43 38 -15 0 0 Z" fill="#063f32"/>
+  <path d="M20 -8 C-58 -107 -87 -107 -116 -159 C-81 -31 -26 -12 20 -8 Z" fill="#063f32" opacity="0.92"/>
+</g>
+
+<g transform="translate(2045 420)">
+  <path d="M0 0 C39 -190 66 -162 86 -238 C60 -53 25 -19 0 0 Z" fill="#0b4b3a"/>
+  <path d="M20 -8 C-39 -133 -58 -133 -78 -198 C-55 -38 -18 -15 20 -8 Z" fill="#0b4b3a" opacity="0.92"/>
+</g>
+
+<g transform="translate(2130 420)">
+  <path d="M0 0 C50 -92 85 -78 110 -115 C78 -26 32 -9 0 0 Z" fill="#155b44"/>
+  <path d="M20 -8 C-50 -64 -75 -64 -100 -96 C-70 -18 -22 -7 20 -8 Z" fill="#155b44" opacity="0.92"/>
+</g>
+
+<g transform="translate(2215 420)">
+  <path d="M0 0 C61 -129 104 -110 134 -161 C95 -36 40 -13 0 0 Z" fill="#092f2a"/>
+  <path d="M20 -8 C-61 -90 -92 -90 -122 -134 C-85 -26 -27 -10 20 -8 Z" fill="#092f2a" opacity="0.92"/>
+</g>
+
+<g transform="translate(2300 420)">
+  <path d="M0 0 C42 -166 71 -141 92 -208 C65 -46 27 -17 0 0 Z" fill="#063f32"/>
+  <path d="M20 -8 C-42 -116 -63 -116 -84 -173 C-59 -33 -19 -13 20 -8 Z" fill="#063f32" opacity="0.92"/>
+</g>
+
+<g transform="translate(2385 420)">
+  <path d="M0 0 C53 -203 90 -173 117 -254 C82 -57 34 -20 0 0 Z" fill="#0b4b3a"/>
+  <path d="M20 -8 C-53 -142 -80 -142 -106 -211 C-74 -41 -24 -16 20 -8 Z" fill="#0b4b3a" opacity="0.92"/>
+</g>
+
+<g transform="translate(2470 420)">
+  <path d="M0 0 C34 -105 58 -89 75 -131 C53 -29 22 -10 0 0 Z" fill="#155b44"/>
+  <path d="M20 -8 C-34 -74 -51 -74 -68 -109 C-48 -21 -15 -8 20 -8 Z" fill="#155b44" opacity="0.92"/>
+</g>
+
+<g transform="translate(2555 420)">
+  <path d="M0 0 C45 -142 76 -121 99 -178 C70 -40 29 -14 0 0 Z" fill="#092f2a"/>
+  <path d="M20 -8 C-45 -99 -68 -99 -90 -148 C-63 -28 -20 -11 20 -8 Z" fill="#092f2a" opacity="0.92"/>
+</g>
+
+<g transform="translate(2640 420)">
+  <path d="M0 0 C56 -179 95 -152 123 -224 C87 -50 36 -18 0 0 Z" fill="#063f32"/>
+  <path d="M20 -8 C-56 -125 -84 -125 -112 -186 C-78 -36 -25 -14 20 -8 Z" fill="#063f32" opacity="0.92"/>
+</g>
+
+<g transform="translate(2725 420)">
+  <path d="M0 0 C37 -81 63 -69 81 -101 C57 -23 24 -8 0 0 Z" fill="#0b4b3a"/>
+  <path d="M20 -8 C-37 -57 -56 -57 -74 -84 C-52 -16 -17 -6 20 -8 Z" fill="#0b4b3a" opacity="0.92"/>
+</g>
+
+<g transform="translate(2810 420)">
+  <path d="M0 0 C48 -118 82 -100 106 -148 C74 -33 31 -12 0 0 Z" fill="#155b44"/>
+  <path d="M20 -8 C-48 -83 -72 -83 -96 -123 C-67 -24 -22 -9 20 -8 Z" fill="#155b44" opacity="0.92"/>
+</g>
+<path d="M0 420 C290 376 574 386 844 405 C1120 425 1434 398 1735 382 C2050 365 2352 392 2800 352 L2800 420 Z" fill="#031d1a" opacity="0.72"/>
+
+<g transform="translate(40.0 110.0) scale(0.920 0.920)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(185.0 140.0) scale(-0.780 0.780)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(315.0 88.0) scale(1.000 1.000)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(455.0 164.0) scale(-0.720 0.720)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(590.0 122.0) scale(0.880 0.880)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(725.0 210.0) scale(-0.620 0.620)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(860.0 74.0) scale(1.040 1.040)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(1015.0 152.0) scale(-0.820 0.820)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(1135.0 111.0) scale(0.940 0.940)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(1285.0 190.0) scale(-0.680 0.680)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(1410.0 126.0) scale(0.860 0.860)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(1545.0 92.0) scale(-0.980 0.980)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(1690.0 170.0) scale(0.720 0.720)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(1815.0 122.0) scale(-0.860 0.860)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(1955.0 72.0) scale(1.050 1.050)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(2115.0 188.0) scale(-0.700 0.700)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(2248.0 136.0) scale(0.840 0.840)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(2385.0 98.0) scale(-0.980 0.980)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(2540.0 164.0) scale(0.760 0.760)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+
+<g transform="translate(2675.0 114.0) scale(-0.900 0.900)">
+  <path d="M0 16 C-8 72 -11 150 -7 250" stroke="#08362e" stroke-width="8" fill="none" stroke-linecap="round"/>
+  <path d="M-5 118 C-55 90 -78 104 -98 142 C-52 151 -27 145 -7 127 Z" fill="#0b4738"/>
+  <path d="M-2 166 C38 132 71 143 94 175 C55 188 25 187 2 174 Z" fill="#155b44"/>
+  <path d="M-8 224 C-44 205 -75 216 -93 244 C-55 257 -28 252 -8 234 Z" fill="#0e4939"/>
+  <g transform="translate(-62 -18) rotate(-8)">
+    <path d="M0 20 L105 0 L138 24 L128 72 L19 88 L-12 62 Z" fill="#f3c747" stroke="#071a33" stroke-width="5" />
+    <path d="M18 84 L128 68 L121 100 L13 113 Z" fill="#0a332b" stroke="#071a33" stroke-width="4" />
+    <path d="M104 30 A28 28 0 1 0 104.1 30" fill="#071a33" />
+    <circle cx="104" cy="30" r="15" fill="#0e4939"/>
+    <circle cx="104" cy="30" r="7" fill="#020a10"/>
+    <circle cx="110" cy="24" r="3.5" fill="#f7dc77"/>
+    <path d="M-12 62 L19 88 L13 113 L-20 86 Z" fill="#0b342e" stroke="#071a33" stroke-width="4" />
+  </g>
+</g>
+</svg>
\ No newline at end of file
diff --git a/images/homepage-face-art-clean.png b/images/homepage-face-art-clean.png
new file mode 100644
index 0000000..8860030
Binary files /dev/null and b/images/homepage-face-art-clean.png differ
diff --git a/images/report/affected-individuals.svg b/images/report/affected-individuals.svg
new file mode 100644
index 0000000..a2d5f76
--- /dev/null
+++ b/images/report/affected-individuals.svg
@@ -0,0 +1,6 @@
+<svg viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="none" stroke="currentColor" stroke-width="1.6" stroke-linecap="round" stroke-linejoin="round">
+  <circle cx="9.8" cy="7.4" r="3.4"/>
+  <path d="M3.6 21c0-4 2.8-6.6 6.2-6.6s6.2 2.6 6.2 6.6"/>
+  <line x1="19.2" y1="4.2" x2="19.2" y2="12.4"/>
+  <circle cx="19.2" cy="16.1" r="0.9" fill="currentColor" stroke="none"/>
+</svg>
diff --git a/images/report/contracts-market.svg b/images/report/contracts-market.svg
new file mode 100644
index 0000000..99c385a
--- /dev/null
+++ b/images/report/contracts-market.svg
@@ -0,0 +1,10 @@
+<svg viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="none" stroke="currentColor" stroke-width="1.6" stroke-linecap="round" stroke-linejoin="round">
+  <path d="M6.2 2.6h8.4l3.2 3.2v15.6H6.2z"/>
+  <path d="M14.6 2.6v3.2h3.2"/>
+  <line x1="8.6" y1="10.4" x2="15.4" y2="10.4"/>
+  <line x1="8.6" y1="13.2" x2="15.4" y2="13.2"/>
+  <line x1="8.6" y1="16" x2="12.4" y2="16"/>
+  <circle cx="16.6" cy="18.2" r="3" fill="#fff3d6"/>
+  <circle cx="16.6" cy="18.2" r="3"/>
+  <path d="M16.6 16.9v2.6M15.7 18.2h1.8"/>
+</svg>
diff --git a/images/report/control-panel.svg b/images/report/control-panel.svg
new file mode 100644
index 0000000..89831d7
--- /dev/null
+++ b/images/report/control-panel.svg
@@ -0,0 +1,9 @@
+<svg viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="none" stroke="currentColor" stroke-width="1.6" stroke-linecap="round" stroke-linejoin="round">
+  <rect x="2.4" y="4" width="19.2" height="14.4" rx="1.2"/>
+  <line x1="6" y1="8" x2="6" y2="14.6"/>
+  <circle cx="6" cy="10.4" r="1.35" fill="currentColor" stroke="none"/>
+  <line x1="12" y1="8" x2="12" y2="14.6"/>
+  <circle cx="12" cy="13" r="1.35" fill="currentColor" stroke="none"/>
+  <line x1="18" y1="8" x2="18" y2="14.6"/>
+  <circle cx="18" cy="9.4" r="1.35" fill="currentColor" stroke="none"/>
+</svg>
diff --git a/images/report/procurement-officials.svg b/images/report/procurement-officials.svg
new file mode 100644
index 0000000..9f30479
--- /dev/null
+++ b/images/report/procurement-officials.svg
@@ -0,0 +1,10 @@
+<svg viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="none" stroke="currentColor" stroke-width="1.6" stroke-linecap="round" stroke-linejoin="round">
+  <path d="M7 4.2h10v17.2H7z"/>
+  <path d="M9.4 2.4h5.2v3.6H9.4z" fill="#fff3d6"/>
+  <path d="M9.4 2.4h5.2v3.6H9.4z"/>
+  <line x1="9.2" y1="11" x2="14.8" y2="11"/>
+  <line x1="9.2" y1="13.6" x2="13.2" y2="13.6"/>
+  <circle cx="16.4" cy="16.2" r="3.2" fill="#fff3d6"/>
+  <circle cx="16.4" cy="16.2" r="3.2"/>
+  <path d="M14.9 16.3l1 1.1 2-2.4"/>
+</svg>
diff --git a/images/report/public-sightings.svg b/images/report/public-sightings.svg
new file mode 100644
index 0000000..dfb4dd7
--- /dev/null
+++ b/images/report/public-sightings.svg
@@ -0,0 +1,6 @@
+<svg viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="none" stroke="currentColor" stroke-width="1.6" stroke-linecap="round" stroke-linejoin="round">
+  <path d="M12 21.6s7-6.4 7-12.4a7 7 0 1 0-14 0c0 6 7 12.4 7 12.4z"/>
+  <rect x="8.8" y="6.6" width="6.4" height="4.6" rx="0.6"/>
+  <circle cx="12" cy="8.9" r="1.4"/>
+  <rect x="10.6" y="5.4" width="1.6" height="1.3" rx="0.3"/>
+</svg>
diff --git a/images/report/system-overview.svg b/images/report/system-overview.svg
new file mode 100644
index 0000000..d234d70
--- /dev/null
+++ b/images/report/system-overview.svg
@@ -0,0 +1,12 @@
+<svg viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="none" stroke="currentColor" stroke-width="1.6" stroke-linecap="round" stroke-linejoin="round">
+  <circle cx="12" cy="12" r="8.4"/>
+  <circle cx="12" cy="12" r="3.1"/>
+  <line x1="12" y1="1.8" x2="12" y2="4.2"/>
+  <line x1="12" y1="19.8" x2="12" y2="22.2"/>
+  <line x1="1.8" y1="12" x2="4.2" y2="12"/>
+  <line x1="19.8" y1="12" x2="22.2" y2="12"/>
+  <line x1="4.7" y1="4.7" x2="6.4" y2="6.4"/>
+  <line x1="17.6" y1="17.6" x2="19.3" y2="19.3"/>
+  <line x1="4.7" y1="19.3" x2="6.4" y2="17.6"/>
+  <line x1="17.6" y1="6.4" x2="19.3" y2="4.7"/>
+</svg>
diff --git a/images/report/test-countermeasures.svg b/images/report/test-countermeasures.svg
new file mode 100644
index 0000000..8a6635a
--- /dev/null
+++ b/images/report/test-countermeasures.svg
@@ -0,0 +1,8 @@
+<svg viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="none" stroke="currentColor" stroke-width="1.6" stroke-linecap="round" stroke-linejoin="round">
+  <path d="M9.5 2.5h5"/>
+  <path d="M10.2 2.5v6.2L4.8 18.4c-.7 1.3.2 2.9 1.7 2.9h11c1.5 0 2.4-1.6 1.7-2.9L13.8 8.7V2.5"/>
+  <path d="M7.6 15.5h8.8"/>
+  <circle cx="10.6" cy="17.6" r="0.55" fill="currentColor" stroke="none"/>
+  <circle cx="13.1" cy="18.4" r="0.55" fill="currentColor" stroke="none"/>
+  <circle cx="12" cy="16.5" r="0.55" fill="currentColor" stroke="none"/>
+</svg>
diff --git a/images/report/weak-spots.svg b/images/report/weak-spots.svg
new file mode 100644
index 0000000..955aa01
--- /dev/null
+++ b/images/report/weak-spots.svg
@@ -0,0 +1,4 @@
+<svg viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="none" stroke="currentColor" stroke-width="1.6" stroke-linecap="round" stroke-linejoin="round">
+  <path d="M12 2.4l7.5 2.7v6.1c0 5-3.2 8.6-7.5 10.4-4.3-1.8-7.5-5.4-7.5-10.4V5.1L12 2.4z"/>
+  <path d="M13.4 6.5l-3.6 5.6h3l-2.4 5.4 5.4-6.6h-3.4l1-4.4z"/>
+</svg>
diff --git a/index.html b/index.html
index d266edb..7720888 100644
--- a/index.html
+++ b/index.html
@@ -1,93 +1,61 @@
 <!doctype html>
-<html lang="it">
+<html lang="en">
 
 <head>
   <meta charset="utf-8" />
   <meta name="viewport" content="width=device-width, initial-scale=1" />
-  <title>Ghòstati</title>
-  <meta name="description" content="Gòstati, diventa fantasma, sfuggi al riconoscimento facciale" />
-
-  <!-- Se mi rompi il cazzo perchè ho messo dei google fonts: hai ragione e fai bene. -->
-  <link rel="preconnect" href="https://fonts.googleapis.com">
-  <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
-
-  <link rel="preconnect" href="https://cdn.jsdelivr.net" crossorigin />
-  <link rel="dns-prefetch" href="https://cdn.jsdelivr.net" />
-
-  <link
-    href="https://fonts.googleapis.com/css2?family=League+Script&family=Outfit:wght@400;600;700;800;900&family=JetBrains+Mono:wght@400;700&display=swap"
-    rel="stylesheet">
-
-  <link rel="stylesheet" href="styles/index.css" />
-  <link rel="stylesheet" href="styles/ghostati.css" />
-
-  <link rel="canonical" href="https://sindacato.nina.watch/ghostati/" />
-  <meta name="robots" content="index, follow, max-image-preview:large">
-
-  <meta property="description"
-    content="Diventa un fantasma, sviluppa e testa camuffamenti anti-riconoscimento biometrico." />
-  <meta property="og:title" content="Ghòstati" />
-  <meta property="og:description"
-    content="Diventa un fantasma, sviluppa e testa camuffamenti anti-riconoscimento biometrico." />
-
-  <meta property="og:type" content="website" />
-  <meta property="og:url" content="https://sindacato.nina.watch/ghostati/" />
-  <meta property="og:image" content="https://sindacato.nina.watch/ghostati/images/ghostati-hero.png" />
-  <meta property="og:site_name" content="NINA" />
-
-  <meta name="twitter:card" content="summary_large_image" />
-  <meta name="twitter:title" content="Ghòstati, col Trucco Avverso" />
-  <meta name="twitter:description"
-    content="Strumento che usa la webcam per insegnarti camuffamenti anti-riconoscimento biometrico - Sviluppa il trucco/plugin!" />
-  <meta name="twitter:image" content="https://sindacato.nina.watch/ghostati/images/ghostati-hero.png" />
-
-  <meta name="theme-color" content="#9f7aea" />
-  <meta name="viewport" content="width=device-width, initial-scale=1.0" />
-  <link rel="apple-touch-icon" href="apple-touch-icon.png" />
-
-  <link rel="icon" type="image/x-icon" href="favicon.ico" />
-  <link rel="sitemap" type="application/xml" title="Sitemap" href="sitemap.xml" />
+  <title>Ghostmaxxing</title>
+  <meta name="description" content="Static homepage concept for Ghostmaxxing, close to the orange editorial mockup." />
+  <link rel="stylesheet" href="styles/pages.css" />
 </head>
 
 <body>
-
-  <header class="ghost-topbar" id="ghostTopbar">
-    <p class="ghost-topbar__progress" id="revealProgress" aria-live="polite">0%</p>
-
-    <button class="ghost-topbar__toggle" id="ghostMenuToggle" type="button" aria-expanded="false"
-      aria-controls="ghostTopbarMenu" aria-label="Apri menu">
-      <span>☰</span>
-    </button>
-
-    <nav class="ghost-topbar__nav" aria-label="Workshop e progetto">
-      <ul class="ghost-topbar__links" id="ghostTopbarMenu">
-        <li><a href="https://github.com/vecna/ghostati" target="_blank" rel="noopener noreferrer">Codice</a></li>
-        <li><a href="https://sindacato.nina.watch/it/iniziative/ghostati/">Cornice politica</a></li>
-        <li><a href="https://ghostyles.vecna.eu">Ghostyles</a></li>
-        <li><a href="docs" target="_blank" rel="noopener noreferrer">Docs</a></li>
-        <li><a href="coverage" target="_blank" rel="noopener noreferrer">
-            Coverage</a></li>
-      </ul>
-    </nav>
-  </header>
-
-  <main class="ghost-hero" id="ghostHero" data-target="ghostati.html">
-    <canvas class="ghost-hero__veil" id="ghostVeil" aria-hidden="true"></canvas>
-
-    <img class="ghost-hero__image" src="images/ghostati-hero.png"
-      alt="Immagine semi-temporanea di un volto generato con un lamerissimo servizio." />
-
-    <a class="ghost-hero__wordmark" id="ghostLink" href="ghostati.html" aria-disabled="true" tabindex="-1">
-      <h1 class="wordmark__layer">Ghòstati!
-        <br>Nessun dato
-        <br>biometrico
-        <br>sarà inviato
-        <br>all'esterno
-      </h1>
-    </a>
+  <main class="editorial-homepage variant-a editorial-homepage--svg-footer-band">
+    <div class="editorial-homepage__shell">
+      <header class="editorial-homepage__header">
+        <a class="wordmark" href="#" aria-label="Ghostmaxxing homepage prototype">
+          <span class="wordmark__title">Ghostmaxxing</span>
+          <span class="wordmark__subtitle">
+            Privacy activism meets public-research platform
+          </span>
+        </a>
+
+        <nav class="editorial-homepage__nav" aria-label="Prototype navigation">
+          <a class="navlink" href="#">Code</a>
+          <a class="navlink" href="about.html">Vision & About</a>
+          <a class="cta-inline" href="report.html">How to <i>leak</i> to us<span aria-hidden="true">↗</span></a>
+        </nav>
+      </header>
+
+      <section class="editorial-homepage__hero" aria-labelledby="hero-title-a">
+        <div class="hero-copy">
+          <!-- <span class="hero-copy__eyebrow">Privacy activist and investigator meets public-research platform</span> -->
+          <h1 class="hero-copy__title" id="hero-title-a">Disrupt<br>the read.</h1>
+          <p class="hero-copy__lead">Ghostmaxxing it's a realtime webapp for testing face-recognition camouflage is used
+            to run <a href="workshops.html">workshops</a>. Runs locally, data stays with you.</p>
+          <p class="hero-copy__small">It can also Augment Reality with overlays over faces, so to drive your makeup with
+            techniques proven to be successful, and <b>Ghostyle</b> is the name of those plugins. <a href="#">Check the
+              Ghostyle archive</a>.</p>
+          <div class="hero-copy__actions">
+            <a class="btn-primary" href="/lab">Open the lab <span aria-hidden="true">↗</span></a>
+          </div>
+          <div class="hero-copy__meta" aria-label="Prototype qualities">
+            <span><a href="/docs">Technical documentation</a></span>
+            <span><a href="/references/">Research References</a></span>
+          </div>
+        </div>
+
+        <figure class="hero-figure">
+          <div class="hero-figure__frame">
+            <img class="hero-figure__image" src="images/ghostmaxxing-homepage.webp"
+              alt="Portrait split between face-recognition overlay and adversarial camouflage." />
+          </div>
+          <!-- <figcaption class="hero-figure__origin">Ghostmaxxing helps you in understanding and mitigating facial recognition technologies.</figcaption> -->
+        </figure>
+      </section>
+    </div>
+    <div class="camera-band" aria-hidden="true"></div>
   </main>
-
-  <script src="scripts/index-effect.js"></script>
 </body>
 
-</html>
+</html>
\ No newline at end of file
diff --git a/ghostati.html b/lab.html
similarity index 94%
rename from ghostati.html
rename to lab.html
index 06b677b..7322c9e 100644
--- a/ghostati.html
+++ b/lab.html
@@ -4,7 +4,7 @@
 <head>
   <meta charset="utf-8" />
   <meta name="viewport" content="width=device-width, initial-scale=1" />
-  <title>Ghòstati</title>
+  <title>Ghostmaxxing</title>
   <!-- Se mi rompi il cazzo perchè ho messo dei google fonts: hai ragione e fai bene. -->
   <link rel="preconnect" href="https://fonts.googleapis.com">
   <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
@@ -14,7 +14,7 @@
 
   <meta name="description"
     content="Laboratorio Web AR per lo sviluppo e il test di trucco anti-riconoscimento biometrico (CV Dazzle)." />
-  <meta property="og:title" content="Ghòstati | Face Lab" />
+  <meta property="og:title" content="Ghostmaxxing | Face Lab" />
   <meta property="og:description"
     content="Laboratorio Web AR in tempo reale per lo sviluppo di camuffamenti anti-riconoscimento biometrico (CV Dazzle) via plugin." />
   <meta property="og:type" content="website" />
@@ -29,6 +29,28 @@
   <link rel="source" href="https://github.com/vecna/ghostati" />
 
   <link rel="stylesheet" href="styles/ghostati.css" />
+
+  <script type="application/ld+json">
+  {
+    "@context": "https://schema.org",
+    "@type": "SoftwareApplication",
+    "@id": "https://sindacato.nina.watch/ghostati/ghostati.html#app",
+    "name": "Ghostmaxxing - Trucco avverso",
+    "applicationCategory": "BrowserApplication",
+    "operatingSystem": "Any",
+    "author": {
+      "@type": "Organization",
+      "@id": "https://www.nina.watch/",
+      "name": "N.I.N.A. - Nè Intelligente Nè Artificiale"
+    },
+    "creator": {
+      "@type": "Person",
+      "@id": "https://me.vecna.eu",
+      "name": "vecna"
+    },
+    "codeRepository": "https://github.com/vecna/ghostati"
+  }
+  </script>
 </head>
 
 <body>
@@ -176,7 +198,7 @@ <h2>Storico volti</h2>
   <!-- Settings Drawer / Menu Overlay -->
   <div id="settingsDrawer" class="settings-drawer hidden">
     <div class="drawer-header">
-      <h2>Ghòstati! Menu</h2>
+      <h2>Ghostmaxxing! Menu</h2>
       <button id="closeSettingsBtn" class="icon-btn" title="Chiudi">
         <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round"
           stroke-linejoin="round">
diff --git a/redesign/brand-voice.txt b/redesign/brand-voice.txt
new file mode 100644
index 0000000..570a721
--- /dev/null
+++ b/redesign/brand-voice.txt
@@ -0,0 +1,571 @@
+# Ghostmaxxing Brand Voice
+
+## Purpose
+
+Ghostmaxxing is a public research lab for testing, explaining, and questioning facial recognition systems through browser-based AR camouflage, adversarial makeup experiments, and curated technical references.
+
+The project should sound like a tool made by people who understand code, research, public space, and political risk — without overselling protection or turning critique into spectacle.
+
+Ghostmaxxing is not a magic invisibility product.
+Ghostmaxxing is a test bench, a visual lab, and a civic intelligence gateway.
+
+## Core voice
+
+Use a tone that is:
+
+- technically precise
+- curious and experimental
+- public-facing but not simplified to the point of being wrong
+- careful with security claims
+- visually imaginative
+- grounded in evidence
+- open to contributors
+- skeptical of biometric inevitability
+- respectful of people whose faces, bodies, and data are affected
+
+The voice is: cyberlab + festival + evidence archive.
+
+Think:
+
+- a public research console
+- a workshop table
+- a browser-based adversarial lab
+- an exhibition label that still respects technical detail
+- a software project that knows it lives in political reality
+
+Avoid sounding like:
+
+- a startup promising disruption
+- a cybersecurity product promising safety
+- an academic paper that excludes non-specialists
+- an activist poster with no technical detail
+- a beauty tutorial with no threat model
+- a hacker toy with no ethics
+
+## One-line description
+
+Ghostmaxxing is a browser-based lab for testing face-recognition camouflage and documenting the technical, cultural, and political history of anti-biometric appearance design.
+
+## Short description
+
+Ghostmaxxing lets people experiment with face-recognition camouflage directly in the browser. It combines webcam-based face detection, modular AR “Ghostyles”, local comparison tests, and a curated reference archive on adversarial makeup, CV Dazzle, face obfuscation, and physical-world attacks against computer vision systems.
+
+## Longer description
+
+Ghostmaxxing is a Web AR laboratory for developing and testing anti-biometric camouflage. It lets developers, researchers, digital rights groups, artists, and curious people apply modular visual overlays to a face, compare recognition behavior before and after the intervention, and understand where this field comes from through a curated technical and cultural references archive.
+
+It does not guarantee anonymity. It does not defeat all facial recognition systems. It helps make biometric claims testable, visible, and discussable.
+
+## Key messages
+
+### 1. Test, do not trust
+
+Facial recognition systems should not be accepted as neutral, inevitable, or universally reliable. Ghostmaxxing makes parts of the recognition pipeline visible enough to test and question.
+
+Suggested phrasing:
+
+- Test the machine before believing the claim.
+- Recognition is not certainty.
+- A face match is a system output, not a truth.
+- Make biometric assumptions inspectable.
+
+Avoid:
+
+- Facial recognition is always broken.
+- This defeats surveillance.
+- This makes you anonymous.
+
+### 2. The face is a technical interface
+
+Ghostmaxxing treats the face as something computer vision systems parse through landmarks, descriptors, regions, contrast, light, occlusion, and learned representations.
+
+Suggested phrasing:
+
+- The face is read as data, geometry, texture, and probability.
+- A Ghostyle is a visual intervention on the machine-readable face.
+- Makeup becomes a test surface.
+- AR becomes a diagnostic layer.
+
+Avoid:
+
+- Your face becomes invisible.
+- This hides your identity.
+- Makeup hacks the system, full stop.
+
+### 3. The references matter
+
+The project is part of a longer genealogy: CV Dazzle, HyperFace, adversarial glasses, adversarial makeup, physical-world optical attacks, NIR attacks, public-space performances, and face-obfuscation research.
+
+Suggested phrasing:
+
+- This field is older, deeper, and more documented than most people think.
+- The references archive maps the technical and cultural lineage of Ghostmaxxing.
+- Each reference asks: what was demonstrated, under what conditions, and what can Ghostmaxxing retest?
+
+Avoid:
+
+- These projects prove facial recognition is useless.
+- Old techniques still work everywhere.
+- A paper result equals real-world protection.
+
+### 4. The whistleblowing/leaking service is central
+
+Ghostmaxxing is a demo but also an attractor of attention
+for people working in that industry, in the search for allies within.
+
+Suggested phrasing:
+
+- Seen facial recognition in public space? Document it and Report it.
+- The lab tests techniques; the reporting node gathers evidence.
+- Technical experiments and civic reporting belong together.
+- Help map where biometric systems are actually used.
+
+Avoid:
+
+- Report everything suspicious.
+- Send us private personal data.
+- Upload faces or sensitive material casually.
+
+### 5. No false safety
+
+This is one of the most important principles.
+
+Suggested phrasing:
+
+- Ghostmaxxing is a research and education tool, not personal protection advice.
+- A failed match in the browser is not a guarantee in the street.
+- Results depend on model, camera, light, pose, distance, and context.
+- Treat every successful evasion as local, conditional, and temporary.
+
+Avoid:
+
+- Protect yourself with Ghostmaxxing.
+- Become unrecognizable.
+- Beat facial recognition.
+- Guaranteed anti-surveillance makeup.
+
+## Vocabulary
+
+### Preferred terms
+
+Use:
+
+- face recognition
+- facial recognition
+- face detection
+- face matching
+- biometric surveillance
+- anti-biometric camouflage
+- adversarial makeup
+- appearance-based intervention
+- physical-world attack
+- computer vision
+- Ghostyle
+- reference archive
+- testability
+- limitations
+- local experiment
+- browser-based lab
+- public research
+- reporting node
+- civic intelligence
+- adversarial
+
+### Use carefully
+
+These can be used, but with context:
+
+- evasion
+- spoofing
+- attack
+- resistance
+- surveillance
+- anonymization
+- camouflage
+
+When using these, specify limits. For example:
+
+Good:
+“a local test of face-matching evasion in a browser pipeline”
+
+Risky:
+“face recognition evasion”
+
+### Avoid or heavily qualify
+
+Avoid:
+
+- invisibility
+- anonymity guarantee
+- protection
+- foolproof
+- defeat
+- bypass all systems
+- anti-surveillance shield
+- safe in public
+- untrackable
+- undetectable
+
+## Naming conventions
+
+### Project name
+
+Preferred:
+
+- Ghòstati
+- Ghostmaxxing, when accents are difficult in filenames, URLs, code, or English-only contexts
+
+Do not over-explain the name every time. Use the accented form in editorial copy and the ASCII-safe form in code contexts.
+
+### Ghostyles
+
+A Ghostyle is a modular visual intervention applied to the face.
+
+Write:
+
+- Ghostyle
+- Ghostyles
+- Ghostyle plugin
+
+Avoid:
+
+- filter, unless speaking to a broad public and immediately clarifying it is not just a cosmetic filter
+- effect, unless describing rendering
+- makeup hack, unless in informal social copy
+
+## Tone by page type
+
+### Homepage
+
+Tone:
+direct, visual, inviting, not too dense.
+
+Goal:
+make people understand what the project is and where to go next.
+
+Good homepage copy:
+
+“Test face-recognition camouflage in the browser. Build Ghostyles, compare detection and matching behavior, and explore the research lineage behind anti-biometric appearance design.”
+
+CTA hierarchy:
+
+1. Open the lab
+2. Explore references
+3. Build a Ghostyle
+4. Report public-space face recognition
+
+### README
+
+Tone:
+technical, precise, contributor-friendly.
+
+Goal:
+help developers understand architecture, run the project, write plugins, run tests, and understand safety limits.
+
+README should include:
+
+- what it does
+- what it does not do
+- architecture
+- local privacy notes
+- install/run commands
+- Ghostyle API
+- tests
+- references page
+- reporting node
+- contribution guidelines
+
+### References page
+
+Tone:
+curated evidence lab.
+
+Goal:
+explain the history, techniques, claims, limitations, and Ghostmaxxing relevance of each reference.
+
+Each reference should answer:
+
+1. What is it?
+2. What did it demonstrate?
+3. What technique did it use?
+4. What are its limits?
+5. Can Ghostmaxxing retest or approximate it?
+6. Where can people read more?
+
+Use concise but substantial cards. Avoid bibliography-only presentation.
+
+### Report / whistleblowing page
+
+Tone:
+careful, calm, protective.
+
+Goal:
+invite useful reports without encouraging unsafe disclosure.
+
+Use:
+
+- “Seen facial recognition in public space?”
+- “Tell us what you know, only if it is safe for you.”
+- “Do not include unnecessary personal data.”
+- “Use the reporting channel that matches your language and risk context.”
+
+Avoid:
+
+- urgent pressure
+- heroic language
+- requests for risky evidence
+- vague claims about secure handling unless verified
+
+### Social posts
+
+Tone:
+sharp, visual, understandable, never overclaiming.
+
+Good pattern:
+
+1. one striking claim
+2. one technical clarification
+3. one action
+
+Example:
+
+“Facial recognition is not magic. It is a pipeline: detection, landmarks, descriptors, matching. Ghòstati lets you test how visual interventions change that pipeline in the browser. Try it, build a Ghostyle, or report public-space deployments.”
+
+### Workshop copy
+
+Tone:
+participatory, practical, safety-aware.
+
+Use:
+
+- “Bring a laptop and curiosity.”
+- “No biometric data is uploaded by the browser lab.”
+- “We will test local recognition behavior, not promise anonymity.”
+- “Failures are useful results.”
+
+## Claims policy
+
+Every public claim should fit one of these categories:
+
+### Safe claim
+
+“Ghostmaxxing lets users test browser-based face detection and matching behavior with visual overlays.”
+
+### Conditional claim
+
+“Some Ghostyles may reduce detection or matching in specific local test conditions.”
+
+### Research claim
+
+“Prior work has shown that physical-world interventions such as makeup, glasses, light, patches, or textiles can affect specific computer vision systems under documented conditions.”
+
+### Unsafe claim
+
+“Ghostmaxxing protects you from facial recognition.”
+
+Do not publish unsafe claims.
+
+## How to describe limitations
+
+Always make limitations explicit when discussing results.
+
+Useful limitation patterns:
+
+- “in this browser pipeline”
+- “under these lighting conditions”
+- “against this model”
+- “with this camera”
+- “in a local comparison test”
+- “not a guarantee against deployed systems”
+- “not tested against newer commercial systems”
+- “not evidence of real-world anonymity”
+
+Example:
+
+“Using this Ghostyle, the browser demo failed to match the saved face under the test conditions. This is a local result, not a general protection claim.”
+
+## Visual language
+
+The visual style should feel like:
+
+- dark public lab
+- neon annotations
+- festival signage
+- accessible technical archive
+- biometric system under inspection
+- face as interface
+- evidence, not spectacle
+
+Good motifs:
+
+- timelines
+- scanlines
+- bounding boxes
+- landmark points
+- masks
+- cards as specimens
+- metadata strips
+- terminal-like labels
+- large CTA buttons
+- subtle glitch, not unreadable glitch
+
+Avoid:
+
+- horror aesthetics
+- military/tactical aesthetics
+- corporate cybersecurity visuals
+- anonymous hacker clichés
+- surveillance fetish imagery
+- beauty branding without critique
+
+## Accessibility
+
+The project can be visually expressive, but must remain readable.
+
+Rules:
+
+- strong contrast
+- legible font sizes
+- mobile-first layout
+- visible focus states
+- buttons that work without hover
+- no essential information only in color
+- no excessive animation
+- reduced-motion support where possible
+- alt text for preview images
+- clear labels for filters and toggles
+
+## Microcopy examples
+
+### Buttons
+
+Use:
+
+- Open the lab
+- Explore references
+- Build a Ghostyle
+- Read the docs
+- View source
+- Report a deployment
+- Switch to genealogy mode
+- Show newest first
+- Filter references
+- Clear filters
+
+Avoid:
+
+- Become invisible
+- Beat surveillance
+- Protect me
+- Hack my face
+
+### Badges
+
+Use:
+
+- Directly testable
+- Partially testable
+- Contextual reference
+- Physical-world
+- Browser lab
+- Face detection
+- Face recognition
+- Visible light
+- Near-infrared
+- Makeup
+- Glasses
+- Textile
+- Patch
+- Survey
+
+### Warning labels
+
+Use:
+
+- Local test only
+- Not a safety guarantee
+- Conditions matter
+- Model-specific result
+- External deployment unknown
+
+## Editorial structure for reference cards
+
+Recommended card order:
+
+1. Year, type, closeness
+2. Title
+3. Authors
+4. Short description
+5. What it demonstrated
+6. Technique / intervention
+7. Target system
+8. Ghostmaxxing testability
+9. Limitations
+10. Links
+
+Recommended labels:
+
+- Demonstrated
+- Why it matters for Ghostmaxxing
+- Ghostmaxxing testability
+- Limits
+- Technique
+- Target
+- Read more
+
+## Writing examples
+
+### Good README sentence
+
+“Ghostmaxxing runs a local browser pipeline for webcam-based face detection, modular AR overlays, and baseline-vs-modified face matching tests.”
+
+### Good limitation sentence
+
+“A browser result can show that a visual intervention affected one local pipeline; it cannot prove protection against deployed facial recognition systems.”
+
+### Good reference sentence
+
+“Adv-Makeup is relevant to Ghostmaxxing because it treats makeup as a physical-world adversarial surface around the face, but its reported effectiveness remains tied to the models and experimental setup used in the paper.”
+
+### Good report CTA
+
+“Seen facial recognition in a public space? Use the reporting node to share what you know, only if it is safe for you to do so.”
+
+## Editorial checklist
+
+Before publishing a new page or section, check:
+
+- Does it say what Ghostmaxxing does technically?
+- Does it avoid protection guarantees?
+- Does it explain limits?
+- Does it link to the lab, repo, docs, references, and reporting node where relevant?
+- Does it help a developer take action?
+- Does it help a non-specialist understand the stakes?
+- Does it avoid collecting or encouraging unnecessary biometric data?
+- Does it use consistent names: Ghostmaxxing, Ghòstati, Ghostyle?
+- Does it preserve the cyberlab + festival tone?
+- Does it make the reporting node visible when discussing real-world deployments?
+
+## Short boilerplate
+
+Use this when space is limited:
+
+“Ghostmaxxing is a browser-based lab for testing face-recognition camouflage. It lets people build and compare AR Ghostyles, explore the research history of anti-biometric appearance design, and report real-world deployments of facial recognition.”
+
+## Safety boilerplate
+
+Use this near demos, workshops, and test results:
+
+“Ghostmaxxing is a research and education tool. Results are local to the tested browser, model, camera, lighting, and conditions. It does not guarantee anonymity or protection against deployed facial recognition systems.”
+
+## References boilerplate
+
+Use this for the references page:
+
+“Ghostmaxxing references archive maps artistic, academic, and activist work around face obfuscation, adversarial makeup, physical-world attacks, and anti-surveillance aesthetics. Each entry documents what was demonstrated, what its limits are, and whether Ghostmaxxing can directly or partially retest the idea.”
+
+## Reporting boilerplate
+
+Use this near the reporting node:
+
+“Technical tests are only one part of the work. If you know of facial recognition being used in public space, the reporting node helps collect evidence about where it is deployed, by whom, with which technology, and under what safeguards or abuses.”
diff --git a/redesign/ghostati-ghostmaxxing-naming-brief.txt b/redesign/ghostati-ghostmaxxing-naming-brief.txt
new file mode 100644
index 0000000..4afc00f
--- /dev/null
+++ b/redesign/ghostati-ghostmaxxing-naming-brief.txt
@@ -0,0 +1,125 @@
+# Ghostati / Ghostmaxxing Naming Brief
+
+## Summary
+
+Ghòstati is the Italian-rooted name.
+Ghostmaxxing is the proposed international-facing name.
+
+The two names can coexist inside the same project universe:
+
+- Ghòstati: local, Italian, NINA-rooted, culturally specific
+- Ghostmaxxing: international, internet-native, easier to explain outside Italy
+- Ghostyle: plugin/style unit, usable across both
+
+## Why Ghostmaxxing works internationally
+
+Ghostmaxxing has several advantages:
+
+1. It is immediately legible to internet-native audiences.
+2. It echoes an existing meme grammar without depending entirely on it.
+3. It suggests “maximizing ghostliness”, which is funny and unexpected.
+4. It keeps the ghost metaphor alive without needing Italian.
+5. It can carry a lighter first impression while the content remains rigorous.
+
+## Risk
+
+The risk is that Ghostmaxxing could sound too unserious.
+
+Counterbalance it with:
+- precise technical copy
+- strong references archive
+- no false anonymity claims
+- clear safety limitations
+- investigative/reporting infrastructure
+
+The meme opens the door; the evidence keeps people inside.
+
+## Recommended public framing
+
+Ghostmaxxing is the international name for Ghòstati: a browser-based lab for testing face-recognition camouflage, building Ghostyles, and documenting the history of anti-biometric appearance design.
+
+## Short versions
+
+Ghostmaxxing is a public lab for testing face-recognition camouflage.
+
+Ghostmaxxing lets people build and test Ghostyles: browser-based visual interventions against face-recognition pipelines.
+
+Ghostmaxxing is not about becoming invisible. It is about making face-recognition systems visible enough to test.
+
+## What not to say
+
+Avoid:
+
+- Ghostmaxxing makes you anonymous.
+- Maximize invisibility.
+- Beat facial recognition.
+- Become untrackable.
+- The meme version of Ghòstati.
+- A fun anti-surveillance filter.
+
+## URL / information architecture options
+
+### Option A: same project, language path
+
+/ghostati/it/ -> Ghòstati
+/ghostati/en/ -> Ghostmaxxing
+
+Pros:
+- clear language split
+- one codebase
+- one project identity
+- easier SEO management
+
+### Option B: same homepage, dual name
+
+Ghòstati / Ghostmaxxing
+
+Pros:
+- immediate continuity
+- no need to explain later
+- useful during transition
+
+Cons:
+- title may feel crowded
+- less clean for international launch
+
+### Option C: Ghostmaxxing landing, Ghòstati as origin
+
+International homepage:
+Ghostmaxxing
+“Originally developed as Ghòstati in the NINA context.”
+
+Pros:
+- clean international presentation
+- preserves origin without overloading the headline
+
+Cons:
+- requires careful explanation to avoid looking like a separate project
+
+## Recommendation
+
+Use Option C for international launch.
+
+Keep Ghòstati as the Italian/local identity and origin.
+Use Ghostmaxxing as the English/international presentation.
+
+## Relationship with Ghostyle
+
+Ghostyle remains a strong term.
+
+It works in both Italian and English:
+- Ghostyle plugin
+- Build a Ghostyle
+- Ghostyle archive
+- Ghostyle test
+
+Ghostmaxxing can become the umbrella action.
+Ghostyle remains the unit of intervention.
+
+Example:
+
+Ghostmaxxing happens through Ghostyles.
+
+Or:
+
+Build a Ghostyle. Test the read. Document the limits.
diff --git a/redesign/visual-direction.txt b/redesign/visual-direction.txt
new file mode 100644
index 0000000..da8545a
--- /dev/null
+++ b/redesign/visual-direction.txt
@@ -0,0 +1,649 @@
+# Ghostati / Ghostmaxxing Visual Direction
+
+## Working definition
+
+Privacy activist and investigator meets public-research platform.
+
+This visual language starts from privacy activism, but it does not sound like policy advocacy. It is investigative, cultural, and public-facing. It wants to reveal what is hidden, document what is unclear, and make biometric systems readable without pretending that visibility alone solves power.
+
+The project is not a neutral tech demo.
+It is not a policy campaign.
+It is not a cybersecurity product.
+It is a public-facing investigative lab about face recognition, camouflage, and the normalization of biometric surveillance.
+
+## Ordered identity labels
+
+Use these labels in this order:
+
+1. activist
+2. cultural / exhibition-like
+3. editorial
+
+Secondary labels:
+
+- investigative
+- public-research
+- anti-biometric
+- design-led
+- technically grounded
+- civic-intelligence oriented
+
+Avoid leading with:
+
+- academic
+- cyberlab
+- advocacy
+- startup
+- security product
+- hacker tool
+- policy campaign
+
+The project can contain academic references, technical tests, and political claims, but the public surface should feel activist first, culturally staged second, editorial third.
+
+## Why not “advocacy”
+
+Avoid “advocacy” as a primary frame.
+
+The tone should not suggest polite policy mediation, institutional lobbying, or reformist optimism. The project may still be useful to journalists, lawyers, researchers, civil society groups, and policy people, but its voice comes from privacy activism and investigation.
+
+Better words:
+
+- investigate
+- reveal
+- document
+- test
+- map
+- contest
+- expose
+- report
+- inspect
+- trace
+- verify
+- archive
+- retest
+
+Use carefully:
+
+- campaign
+- public-interest
+- civic
+- rights
+- resistance
+
+Avoid as core identity:
+
+- advocate
+- awareness campaign
+- policy solution
+- compliance
+- responsible innovation
+- trust framework
+
+## Core metaphor
+
+Surveillance has become part of the landscape.
+
+It grows everywhere.
+It is installed as infrastructure.
+It is normalized as safety, convenience, and administration.
+It becomes background scenery before people have understood who operates it, what it sees, what it stores, and who can access it.
+
+The “camera flowers” motif captures this:
+from a distance, the bottom of the page looks decorative, almost botanical.
+On closer inspection, the flowers are cameras.
+
+This is the central visual metaphor:
+
+> Surveillance is naturalized as landscape.
+
+The design should make that naturalization feel beautiful, unsettling, and legible.
+
+## Short style description
+
+Privacy activist and investigator meets public-research platform.
+
+A more visual phrase:
+
+Activist cultural poster + investigative public lab.
+
+A more refined phrase:
+
+Design-led privacy activism with an investigative archive sensibility.
+
+A more international phrase:
+
+An activist, exhibition-like interface for investigating face recognition.
+
+## What this style is
+
+- activist
+- culturally staged
+- investigative
+- editorial
+- public-facing
+- visually composed
+- technically literate
+- symbolic
+- slightly formal
+- direct without being sloganistic
+- beautiful but not decorative-only
+- unsettling without horror aesthetics
+
+## What this style is not
+
+- cyberpunk
+- dark terminal dashboard
+- startup SaaS
+- academic repository
+- policy NGO template
+- hacker conference visual identity
+- corporate cybersecurity
+- beauty brand
+- paranoid survival guide
+- streetwear-only meme aesthetic
+- military/tactical interface
+
+## Visual mood
+
+The mood is:
+
+- bold
+- warm
+- public
+- theatrical
+- investigative
+- composed
+- slightly dissonant
+- attractive at first glance
+- stranger on second glance
+
+The page should make people feel:
+“this is beautiful”
+then:
+“wait, those flowers are cameras”
+then:
+“this is about how surveillance becomes ordinary”
+
+## Color direction
+
+Primary palette:
+
+- hot orange background
+- black serif headlines
+- dark navy body text
+- green and yellow botanical/camera illustration
+- occasional cream or pale yellow accent
+- occasional deep green shadow
+
+Suggested palette names, not strict values:
+
+- surveillance orange
+- ink black
+- institutional navy
+- leaf green
+- warning yellow
+- faded cream
+
+Example CSS tokens:
+
+:root {
+  --gm-bg: #f26a1b;
+  --gm-bg-deep: #df5512;
+  --gm-ink: #050505;
+  --gm-text: #071a33;
+  --gm-muted: #253a54;
+  --gm-green: #063f32;
+  --gm-green-soft: #1e6b4f;
+  --gm-yellow: #f3c747;
+  --gm-cream: #ffe7a8;
+  --gm-border: rgba(5, 5, 5, 0.28);
+}
+
+The orange should feel like a poster wall, a public notice, a heat field, or institutional alarm color — not neon cyberpunk.
+
+## Typography
+
+### Headlines
+
+Use a strong black serif display face.
+
+Qualities:
+- high contrast
+- editorial
+- elegant but assertive
+- large enough to dominate
+- more poster than interface
+
+Good feeling:
+- cultural magazine
+- exhibition title
+- public campaign poster
+- manifesto without shouting
+
+Avoid:
+- tech sans headlines
+- glitch fonts
+- fake terminal fonts
+- overly nostalgic typewriter styles
+
+### Body text
+
+Use dark navy, not pure black, for body copy.
+
+Qualities:
+- readable
+- contemporary
+- clean
+- structured
+- precise
+
+Body text should not feel corporate. It should feel like careful public explanation.
+
+### Metadata
+
+Small metadata can use a clean sans or monospace sparingly, but avoid letting the whole page become technical-HUD.
+
+Use monospace only for:
+- years
+- tags
+- reference metadata
+- short labels
+- testability badges
+
+## Image language
+
+The key image language is contrast between:
+
+1. machine-readable face
+2. adversarial / ornamental / human intervention
+3. surveillance as naturalized landscape
+
+### Face treatment
+
+The face can be split:
+
+- one side: face-recognition overlay, landmarks, bounding box, confidence markers
+- other side: adversarial makeup, dazzle pattern, organic camouflage
+
+Keep it elegant, not sci-fi.
+
+The recognition overlay should be present enough to be legible, but not dominate the aesthetic. Avoid excessive HUD elements, scanline overload, or military target visuals.
+
+### Camera flowers
+
+The bottom illustration should be SVG-like and aligned to the bottom of the page.
+
+It should resemble a field of flowers at first glance:
+- stems
+- leaves
+- repeated shapes
+- irregular heights
+- botanical rhythm
+
+But the blossoms are:
+- CCTV cameras
+- sensor boxes
+- camera housings
+- small lenses
+
+Use green and yellow on orange.
+
+The motif should be reusable across pages:
+- landing footer
+- report page divider
+- references header detail
+- social cards
+- stickers/posters
+
+## Interaction
+
+No animation is required.
+
+The mockup proves that the idea works through composition, title, claim, and image.
+
+Use animation only when it clarifies the concept or creates a small intentional gesture. Do not depend on it to make the landing understandable.
+
+For the landing:
+
+Preferred:
+- static, immediate CTA
+- strong hero image
+- clear title
+- optional subtle hover/mouse reveal
+- no mandatory unlock mechanism
+
+Avoid:
+- hidden primary link
+- long reveal interaction
+- friction that blocks access
+- interaction needed to understand the page
+
+A small interaction can remain as an enhancement:
+- hover reveals more makeup
+- pointer movement shifts recognition/camouflage balance
+- reduced-motion users see final state immediately
+
+But the page should work perfectly as a poster.
+
+## Landing page direction
+
+### Primary headline
+
+Strong options:
+
+- Disrupt the read.
+- Make the read visible.
+- The face is not a truth.
+- Recognition is a system output.
+- Test the face-reading machine.
+- Camouflage the pipeline.
+
+The current strongest option:
+
+Disrupt the read.
+
+It is short, technical enough, and visually powerful. It does not overclaim.
+
+### Supporting claim
+
+Use:
+
+Ghostati tests face-recognition camouflage through browser-based overlays and local experiments. Knowledge stays with you.
+
+For the international version:
+
+Ghostmaxxing tests face-recognition camouflage through browser-based overlays, local experiments, and a public archive of anti-biometric techniques.
+
+### CTA hierarchy
+
+Primary:
+- Open the lab
+
+Secondary:
+- Explore references
+- Report a deployment
+- View source
+
+Avoid:
+- Become invisible
+- Beat surveillance
+- Protect yourself
+- Hack your face
+
+### Layout
+
+- top nav with wordmark
+- large serif headline on the left
+- face image on the right or centered-right
+- body text below headline
+- primary CTA button
+- secondary text link
+- camera flowers anchored at bottom
+- generous whitespace
+- strong orange field
+
+The page should feel like a poster that happens to be interactive.
+
+## References page direction
+
+The references page should inherit the new style but become more structured.
+
+Use:
+- orange or cream background sections
+- black serif titles
+- dark navy explanatory text
+- card layout
+- tags in navy/green/yellow
+- camera-flower motif in header or footer
+- less HUD, more archive
+
+The references page is not a cyberlab database.
+It is an investigative archive.
+
+Each card should answer:
+- What is this?
+- What did it demonstrate?
+- Why does it matter?
+- Can Ghostati/Ghostmaxxing retest it?
+- What are the limitations?
+- Where can I read more?
+
+## Report / whistleblowing page direction
+
+This page should feel calm, careful, and investigative.
+
+Use the same color system, but reduce spectacle.
+
+Tone:
+- protective
+- precise
+- serious
+- not heroic
+- not paranoid
+- not urgent in a manipulative way
+
+Key claim:
+
+Seen facial recognition in public space? Help document where it appears, who operates it, what technology is used, and what safeguards or abuses are present.
+
+Use the camera flowers motif more literally here:
+the surveillance landscape is what the reporting node tries to map.
+
+## App / Ghostati interface direction
+
+The actual lab interface can be more technical than the landing, but should not revert entirely to cyberlab aesthetics.
+
+Use the new visual direction as a layer:
+
+- dark navy panels instead of pure black cyber panels
+- orange/yellow accents
+- serif only for major labels or title areas
+- clean sans for controls
+- recognition overlays can remain technical
+- Ghostyle overlays can be more ornamental
+- status language should stay precise
+
+The lab can still look like a tool.
+But it should belong to the same world as the landing.
+
+## Naming architecture
+
+### Italian context
+
+Use:
+
+Ghòstati
+
+Why:
+- works as a direct imperative in Italian
+- carries the “become ghost / stay ghosted” resonance
+- fits the local NINA context
+- feels rooted and specific
+
+Italian pages can remain more locally contextual and speak to the audience already explored:
+- public-space deployments
+- Italian legal/political context
+- NINA reporting node
+- workshop/festival language
+
+### International context
+
+Use:
+
+Ghostmaxxing
+
+Why:
+- meme-aware and contemporary
+- immediately suggests “becoming more ghost-like”
+- funny and unexpected
+- legible to internet-native audiences
+- less dependent on Italian wordplay
+- can carry a lighter public-entry tone without weakening the technical work
+
+Ghostmaxxing should not become unserious. The meme should open the door; the project should remain rigorous.
+
+### Relationship between names
+
+Recommended architecture:
+
+- Ghòstati: Italian/local name and original project root
+- Ghostmaxxing: international-facing name for the same project universe
+- Ghostyle: plugin/style unit, usable in both contexts
+
+Possible wording:
+
+Ghostmaxxing is the international name for Ghòstati: a browser-based lab for testing face-recognition camouflage and documenting anti-biometric appearance design.
+
+Or:
+
+Ghòstati / Ghostmaxxing is a public lab for testing how faces become machine-readable — and how that reading can be disrupted.
+
+### Avoid confusion
+
+Do not present Ghostmaxxing as a separate fork unless it actually becomes one.
+
+Use:
+- “international name”
+- “international-facing version”
+- “English-language presentation”
+- “same project universe”
+
+Avoid:
+- “rebrand” too early
+- “new project” unless architecture changes
+- “meme project”
+- “funny version”
+
+## Ghostmaxxing tone
+
+Ghostmaxxing can be slightly more internet-native, but not unserious.
+
+It can use:
+- sharper headlines
+- cleaner English
+- clearer CTAs
+- a bit more playfulness
+- stronger visual identity
+
+It should still avoid:
+- anonymity promises
+- overclaiming
+- “beat face recognition”
+- meme overload
+- ironic detachment
+
+Good phrase:
+
+Ghostmaxxing is not about becoming invisible. It is about making face-recognition systems visible enough to test.
+
+## Copy examples
+
+### Landing
+
+Disrupt the read.
+
+Ghostmaxxing tests face-recognition camouflage through browser-based overlays and local experiments. Explore the archive, build a Ghostyle, or report where facial recognition appears in public space.
+
+### References
+
+The archive maps artistic, technical, and activist work around face obfuscation, adversarial makeup, and physical-world attacks on computer vision systems. Each entry explains what was demonstrated, what remains limited, and whether the idea can be retested in Ghostmaxxing.
+
+### Report page
+
+Seen facial recognition in public space? Help document where it appears, who operates it, what technology is used, and what safeguards or abuses are present. Share only what is safe for you to share.
+
+### App disclaimer
+
+Local tests are not protection claims. Results depend on model, camera, light, pose, distance, and context.
+
+## Design principles
+
+### 1. Beauty first, unease second
+
+The page should attract before it disturbs.
+
+The flowers should be beautiful. Then the viewer notices they are cameras.
+
+### 2. No false safety
+
+Never imply that a Ghostyle makes someone safe from surveillance.
+
+### 3. Make systems visible
+
+The visual language should reveal pipelines, infrastructures, and hidden arrangements.
+
+### 4. Less cyber, more public
+
+Avoid the cliché of surveillance represented only through dark screens and green grids.
+
+Surveillance is not just in computers. It is in streets, schools, stadiums, shops, airports, phones, buildings, and public administration.
+
+### 5. The archive is part of the interface
+
+References are not an appendix.
+They are evidence of lineage, repetition, and accumulated knowledge.
+
+### 6. Investigation over advocacy
+
+The project should invite people to look, test, report, and verify — not merely agree.
+
+## Visual checklist
+
+Before publishing a page, check:
+
+- Does it feel activist first?
+- Does it have cultural/exhibition quality?
+- Does it feel editorially composed?
+- Does it avoid old cyberlab tropes?
+- Does it use strong contrast?
+- Does it include the camera-as-landscape metaphor when appropriate?
+- Is the CTA obvious?
+- Is the page readable without animation?
+- Are claims precise and limited?
+- Is the reporting node visible where it matters?
+- Would the design still work as a poster?
+- Would the page make sense to a developer, an investigator, and a curious person?
+
+## Practical CSS direction
+
+Use global design tokens for the new visual system:
+
+:root {
+  --gm-bg: #f26a1b;
+  --gm-bg-deep: #df5512;
+  --gm-ink: #050505;
+  --gm-text: #071a33;
+  --gm-muted: #253a54;
+  --gm-green: #063f32;
+  --gm-green-soft: #1e6b4f;
+  --gm-yellow: #f3c747;
+  --gm-cream: #ffe7a8;
+  --gm-border: rgba(5, 5, 5, 0.28);
+  --gm-panel: rgba(255, 231, 168, 0.16);
+}
+
+Typography direction:
+
+- display serif for h1/h2/wordmark
+- readable sans for body and nav
+- small monospace only for tags, years, and metadata
+
+Layout direction:
+
+- poster-scale hero
+- clear CTA cluster
+- bottom-aligned illustrated motif
+- large type on desktop
+- compressed but still dramatic type on mobile
+
+## Future tasks
+
+1. Build a static SVG camera-flower border.
+2. Replace the cyberlab landing prototype with a static editorial landing prototype.
+3. Update brand-voice.txt to include Ghostmaxxing and the new visual direction.
+4. Adapt references.css away from cyberlab toward investigative archive.
+5. Create a report.html design using the same visual system.
+6. Define international copy around Ghostmaxxing.
+7. Decide whether Ghostati and Ghostmaxxing share one homepage or separate language-specific entry points.
+8. Create social cards using the orange/camera-flower/serif system.
+9. Build reusable CSS tokens and components for the broader site.
diff --git a/references/NOTES.md b/references/NOTES.md
new file mode 100644
index 0000000..4d9bbde
--- /dev/null
+++ b/references/NOTES.md
@@ -0,0 +1,67 @@
+# References toolkit — maintenance notes
+
+`references.html` is generated from `REFERENCES.json`. Three small command-line
+tools do the work: two are plain Node (nothing to install), one is Python and
+needs a virtual environment because it talks to the real internet.
+
+None of these run automatically — you run each one by hand, in this order,
+whenever you add or edit an entry.
+
+## The three tools
+
+### 1. `scripts-dev/validate-references.js`
+Checks `REFERENCES.json` against the rules the schema and the build script
+both assume: required fields present, enum values valid (`type`,
+`ghostati_testability.level`, ...), tags used in `intervention`/`target`/
+`domain` are declared in `tag_definitions`, no duplicate slugs or titles.
+
+- **Reads:** `REFERENCES.json`
+- **Writes:** nothing. Prints errors/warnings to the console and exits
+  non-zero on failure.
+- **Run:**
+  ```bash
+  node scripts-dev/validate-references.js
+  ```
+
+Run this first after any manual or LLM-assisted edit to the JSON — it's the
+thing that would have caught the `ghostmaxxing_relevance` vs. `ghostati_relevance`
+field-name mismatch that was in the old prompt file.
+
+### 2. `scripts-dev/build-references-page.js`
+Regenerates the reference list itself. Calls the validator automatically
+first and refuses to build if it fails.
+
+- **Reads:** `REFERENCES.json`, `templates/references.template.html`
+- **Writes:** `references.html`
+- **Run:**
+  ```bash
+  node scripts-dev/build-references-page.js
+  ```
+
+The header, nav, intro copy, and the "How this archive is maintained"
+section live in `templates/references.template.html` — edit that file
+directly for wording changes. The script only fills in its `{{...}}`
+placeholders (`{{REFERENCE_COUNT}}`, `{{STATS_LINE}}`, `{{DECADE_NAV}}`,
+`{{REFERENCES_BODY}}`); it never touches the rest of that file.
+
+### 3. PROMPT-for-REFERENCES-metadata
+
+
+1. Edit `REFERENCES.json` by hand, or generate/extend it with an LLM using
+   `PROMPT-REFERENCES-UPDATE.txt`.
+2. `node scripts-dev/validate-references.js` — fix anything it flags.
+3. Get the preview image in place:
+   - Has a paper/arXiv link → `python3 scripts-dev/extract-paper-figures.py --slug <slug>`,
+     review the candidate, move/rename it to `images/references/<slug>.png`.
+   - No paper (art/activism entry) → source or make a PNG by hand, save it
+     to `images/references/<slug>.png`.
+4. Confirm `preview_image` in the JSON points at `/images/references/<slug>.png`.
+5. `node scripts-dev/build-references-page.js` to regenerate `references.html`.
+6. Open `references.html` in a browser and spot-check the new card.
+
+
+After generating or editing REFERENCES.json:
+
+1. Run `node scripts-dev/validate-references.js` and fix anything it flags — it checks required fields, enum values, and tag_definitions consistency, and will catch a wrong field name (e.g. ghostati_relevance vs. ghostmaxxing_relevance) before it becomes a silent bug in the build.
+2. Get preview_image assets in place per the Preview-image rules above.
+3. Run `node scripts-dev/build-references-page.js` to regenerate references.html. It calls the validator itself and refuses to build on a failing JSON, but running it manually first gives you a clearer look at any errors.
diff --git a/references/PROMPT-for-REFERENCES-metadata.txt b/references/PROMPT-for-REFERENCES-metadata.txt
new file mode 100644
index 0000000..766e4d5
--- /dev/null
+++ b/references/PROMPT-for-REFERENCES-metadata.txt
@@ -0,0 +1,125 @@
+Produce or extend a JSON reference set for the same project universe as the existing file "REFERENCES.json".
+
+Goal:
+Maintain a coherent, deduplicated set of cultural, artistic, academic, and activism-adjacent references around:
+- anti-surveillance aesthetics
+- face obfuscation / face camouflage
+- makeup-based or appearance-based interventions
+- adversarial attacks on face recognition
+- physical-world adversarial vision techniques adjacent to facial recognition
+- sensor-specific attacks that help explain why different capture technologies require different countermeasures
+
+Output rules:
+1. Output valid JSON only.
+2. Preserve the top-level structure:
+{
+  "schema_version": "2.1",
+  "project_scope": "...",
+  "ordering_definition": "...",
+  "language": "en",
+  "closeness_definition": {...},
+  "tag_definitions": {...},
+  "references": [...]
+}
+3. Each reference object MUST contain at least:
+- slug
+- title
+- author
+- year
+- link
+- links
+- type
+- closeness
+- description
+- demonstrated
+- ghostati_relevance
+- ghostati_testability
+- intervention
+- target
+- domain
+- limitations
+- reproducibility
+- citation
+4. All editorial text MUST be written in English.
+5. Preserve existing slugs for existing references.
+6. Return the full updated JSON, not just the new or changed entries.
+
+Field conventions:
+- slug: kebab-case unique identifier stable across updates.
+- title: canonical public title of the work/project/paper.
+- author: array of author/creator names in display order.
+- year: integer year of publication, release, first public presentation, or first documented appearance.
+- period: optional human-readable range, for projects with a multi-year lifespan such as "2019–2021".
+- link: stable canonical link preferred in this order: DOI > arXiv abs page > official project page > official institutional page.
+- links: object with canonical, paper, project, code, video, doi, arxiv. Use null for unknown or unavailable fields.
+- type: exactly one of "artistic" | "research" | "activism".
+- closeness: integer from 1 to 100, not a string and no percent sign.
+- description: short editorial description for the public references page, 1–2 sentences.
+- demonstrated: concise technical or curatorial summary of what the work actually showed, built, tested, or performed.
+- ghostati_relevance: why this reference matters to Ghostmaxxing specifically.
+- ghostati_testability: object with:
+  - level: exactly one of "direct" | "partial" | "indirect" | "contextual" | "no"
+  - reason: short explanation of what Ghostmaxxing can or cannot retest.
+- intervention: array of normalized tags describing the technique or practice.
+- target: array of normalized tags describing the system or pipeline stage affected.
+- domain: array of normalized tags describing the operational domain.
+- limitations: explicit limitations. Use stated limitations when available; infer cautiously when needed. Do not overclaim.
+- reproducibility: short normalized phrase such as "paper-available", "code-available", "documented-project", "artwork-documentation", "survey", "unclear".
+- citation: short display citation such as "Yin et al., IJCAI 2021" or "Adam Harvey, CV Dazzle, 2010".
+- visual_hint: optional short internal hint used to generate or select a preview image.
+
+Closeness rubric:
+- 100 = exact anchor cluster for this project world:
+  * Michelle Tylicki + Lauri Love — DAZZLE
+  * Adam Harvey — CV Dazzle
+- 90–99 = direct face-centered anti-surveillance / makeup / camouflage references.
+- 70–89 = directly relevant face-recognition attack literature or highly adjacent face-obfuscation work.
+- 40–69 = enabling or adjacent adversarial-vision methods that matter conceptually but are less face/makeup-specific.
+- 1–39 = peripheral or modality-shifted references, especially infrared/night-vision/sensor-jamming examples.
+
+Ghostmaxxing testability rubric:
+- direct: Ghostmaxxing can retest a closely similar browser-visible face intervention or matching/detection behavior.
+- partial: Ghostmaxxing can approximate one part of the intervention, but not the full experimental setup.
+- indirect: Ghostmaxxing can use the reference conceptually, but needs different sensors, models, or optimization to validate it.
+- contextual: the reference helps explain history, public practice, aesthetics, or methodology, but is not directly testable.
+- no: the reference is relevant to the wider map but cannot be tested with Ghostmaxxing’s browser AR pipeline.
+
+Tag rules:
+- Tags MUST be lowercase kebab-case.
+- Prefer existing tags from tag_definitions.
+- If adding a new tag, also add it to tag_definitions.
+- Do not create near-duplicates such as "face-recognition" and "facial-recognition".
+
+Research and limitation rules:
+- Read the canonical source before adding or changing demonstrated, limitations, links, or testability.
+- Use primary sources where possible: DOI page, arXiv abstract page, official project page, official repository, institutional page.
+- Do not invent missing metadata.
+- If limitations are not explicit, infer conservatively and phrase them as scope limits.
+- Good limitation pattern: "The reported result depends on the tested model family, acquisition setup, physical realization, and subjects. Effectiveness against newer production systems is not established by this reference."
+- Avoid: "This works", "This defeats facial recognition", "This proves protection".
+
+Normalization rules:
+- Deduplicate by normalized title first, then by DOI or canonical link.
+- Prefer canonical capitalization.
+- Prefer full author lists when available.
+- Keep the existing entry if the new candidate is the same work with a worse link.
+- If the same work exists with a better canonical link, replace only the link and preserve closeness unless there is a strong reason to revise it.
+- Preserve the existing slug whenever the same work reappears.
+
+Ordering logic:
+- Return the full updated JSON sorted newest first:
+  year descending, closeness descending, title ascending.
+- references.html is generated newest-first, grouped into decade sections; there is no interactive toggle to reverse the order (the old filter/genealogy-mode UI was removed — see scripts-dev/build-references-page.js).
+
+Style constraints:
+- Avoid generic press coverage unless it is itself the primary artistic or activist source.
+- Prefer official project pages, DOI links, arXiv abs pages, institutional pages, or artist pages.
+- Keep the set operational: every added item should help situate the project culturally, artistically, academically, or politically.
+- Descriptions should be concise, readable, and suitable for a static references page.
+- Every entry should help answer:
+  1. What is it?
+  2. What did it demonstrate?
+  3. What technique did it use?
+  4. What are its limits?
+  5. Can Ghostmaxxing directly or partially retest it?
+  6. Where can people read more?
diff --git a/references/REFERENCES.json b/references/REFERENCES.json
new file mode 100644
index 0000000..adff3e6
--- /dev/null
+++ b/references/REFERENCES.json
@@ -0,0 +1,1061 @@
+{
+  "schema_version": "2.1",
+  "project_scope": "anti-surveillance / face-obfuscation / makeup-adjacent artistic, research, and activism references",
+  "ordering_definition": "Default rendering order is newest first. Primary key year descending; secondary key closeness descending; tertiary key title ascending. The HTML page also exposes a genealogy mode that reverses the year order.",
+  "language": "en",
+  "closeness_definition": {
+    "description": "Integer 1-100 indicating how close a source is to the conceptual core of the project: face-centered anti-surveillance through appearance design, makeup, camouflage, or directly adjacent face-recognition attacks.",
+    "anchor_examples": {
+      "100": [
+        "Michelle Tylicki + Lauri Love — DAZZLE",
+        "Adam Harvey — CV Dazzle"
+      ],
+      "20": [
+        "infrared / night-vision sensor disturbance references such as The Camera-Shy Hoodie"
+      ]
+    },
+    "bands": {
+      "90_100": "direct face-centered anti-surveillance / makeup / camouflage references",
+      "70_89": "directly relevant face-recognition attack literature or strongly adjacent face-obfuscation work",
+      "40_69": "enabling or adjacent adversarial-vision methods",
+      "1_39": "peripheral or modality-shifted references, especially infrared / sensor-jamming / non-face-centric work"
+    }
+  },
+  "tag_definitions": {
+    "type": [
+      "artistic",
+      "research",
+      "activism"
+    ],
+    "ghostati_testability_level": [
+      "direct",
+      "partial",
+      "indirect",
+      "contextual",
+      "no"
+    ],
+    "intervention": [
+      "adversarial-ml",
+      "adversarial-optimization",
+      "digital-perturbation",
+      "decoy",
+      "environmental-perturbation",
+      "face-obfuscation",
+      "false-face",
+      "garment",
+      "glasses",
+      "hair",
+      "infrared",
+      "interactive-tool",
+      "light",
+      "makeup",
+      "near-infrared",
+      "optical",
+      "orbital-region",
+      "patch",
+      "pattern",
+      "performance",
+      "printed",
+      "projection",
+      "public-walk",
+      "representation-attack",
+      "sensor-aware-material",
+      "sensor-jamming",
+      "semantic-perturbation",
+      "shadow",
+      "structured-light",
+      "survey",
+      "taxonomy",
+      "textile",
+      "toolkit",
+      "wearable-accessory",
+      "workshop",
+      "asymmetry"
+    ],
+    "target": [
+      "3d-face-recognition",
+      "certified-defense",
+      "deep-representation",
+      "depth-face-recognition",
+      "face-detection",
+      "face-recognition",
+      "face-verification",
+      "image-classification",
+      "machine-learning-models",
+      "nir-face-recognition",
+      "object-detection",
+      "public-space-surveillance",
+      "security-camera",
+      "sensor-disturbance",
+      "traffic-sign-recognition"
+    ],
+    "domain": [
+      "3d",
+      "browser",
+      "digital",
+      "environmental",
+      "festival",
+      "near-infrared",
+      "optical",
+      "physical-world",
+      "printed",
+      "public-space",
+      "software-library",
+      "visible-light",
+      "wearable"
+    ]
+  },
+  "references": [
+    {
+      "slug": "human-imperceptible-nir-face-attack",
+      "title": "Human-Imperceptible Physical Adversarial Attack for NIR Face Recognition Models",
+      "author": [
+        "Songyan Xie",
+        "Jinghang Wen",
+        "Encheng Su",
+        "Qiucheng Yu"
+      ],
+      "year": 2025,
+      "link": "https://arxiv.org/abs/2504.15823",
+      "type": "research",
+      "closeness": 22,
+      "description": "A 2025 physical attack on near-infrared face recognition using infrared-absorbing ink and optimized patches. It is peripheral to Ghostati’s visible-makeup workflow, but important for understanding how anti-biometric interventions move toward sensors and spectra that humans cannot see.",
+      "visual_hint": "nir_subtle_points",
+      "demonstrated": "The paper demonstrates a black-box physical adversarial patch for NIR face recognition, using infrared-absorbing ink, skin-reflection modeling, and optimized patch placement to improve attack success in digital and physical domains.",
+      "ghostati_relevance": "It extends the anti-recognition design space beyond visible cosmetics into NIR-specific materials. Ghostati can use it as a conceptual reference for sensor-aware camouflage, but cannot directly reproduce NIR behavior with a normal browser webcam pipeline.",
+      "ghostati_testability": {
+        "level": "indirect",
+        "reason": "Ghostati can document the concept of localized face-region perturbations, but it cannot verify NIR absorption, NIR imaging, or infrared-specific recognition models without dedicated hardware and models."
+      },
+      "intervention": [
+        "infrared",
+        "patch",
+        "sensor-aware-material"
+      ],
+      "target": [
+        "nir-face-recognition",
+        "face-verification"
+      ],
+      "domain": [
+        "physical-world",
+        "near-infrared",
+        "wearable"
+      ],
+      "limitations": "The reported results are tied to NIR cameras, tested models, patch materials, subject set, and controlled acquisition conditions. Effectiveness against visible-light systems, production deployments, or newer NIR pipelines is not established by this reference.",
+      "reproducibility": "artifact-available",
+      "citation": "Xie et al., arXiv 2025",
+      "links": {
+        "canonical": "https://arxiv.org/abs/2504.15823",
+        "paper": "https://arxiv.org/abs/2504.15823",
+        "project": null,
+        "code": null,
+        "video": null,
+        "doi": "https://doi.org/10.48550/arXiv.2504.15823",
+        "arxiv": "https://arxiv.org/abs/2504.15823"
+      }
+    },
+    {
+      "slug": "accessorize-in-the-dark",
+      "title": "Accessorize in the Dark: A Security Analysis of Near-Infrared Face Recognition",
+      "author": [
+        "Amit Cohen",
+        "Mahmood Sharif"
+      ],
+      "year": 2024,
+      "link": "https://doi.org/10.1007/978-3-031-51479-1_3",
+      "type": "research",
+      "closeness": 22,
+      "description": "A security analysis of near-infrared face recognition that adapts adversarial accessories to NIR-based systems used in access-control and high-stakes settings.",
+      "visual_hint": "nir_points",
+      "demonstrated": "The work tests NIR-based face recognition against physically realizable adversarial accessories, reporting high physical attack success across defended and undefended models.",
+      "ghostati_relevance": "It is relevant because it shows that face-recognition fragility is not limited to visible-light cameras or RGB makeup. It helps Ghostati explain why different sensors require different camouflage strategies.",
+      "ghostati_testability": {
+        "level": "partial",
+        "reason": "Ghostati can approximate the visible placement of a wearable accessory around the face, but cannot test NIR capture, NIR-specific perturbations, or defended NIR models in the browser."
+      },
+      "intervention": [
+        "glasses",
+        "near-infrared",
+        "wearable-accessory"
+      ],
+      "target": [
+        "nir-face-recognition",
+        "face-verification"
+      ],
+      "domain": [
+        "physical-world",
+        "near-infrared",
+        "wearable"
+      ],
+      "limitations": "The result depends on NIR capture assumptions, model set, accessory realization, and experimental protocol. A standard RGB webcam test in Ghostati would not confirm the NIR-specific claim.",
+      "reproducibility": "code-available",
+      "citation": "Cohen and Sharif, ESORICS 2023 / LNCS 2024",
+      "links": {
+        "canonical": "https://doi.org/10.1007/978-3-031-51479-1_3",
+        "paper": "https://link.springer.com/chapter/10.1007/978-3-031-51479-1_3",
+        "project": null,
+        "code": "https://github.com/AmitCohen3/Accessorize-in-the-dark",
+        "video": null,
+        "doi": "https://doi.org/10.1007/978-3-031-51479-1_3",
+        "arxiv": null
+      }
+    },
+    {
+      "slug": "dazzle",
+      "title": "DAZZLE",
+      "author": [
+        "Michelle Tylicki",
+        "Lauri Love"
+      ],
+      "year": 2023,
+      "link": "https://www.michelletylicki.info/dazzle/",
+      "type": "activism",
+      "closeness": 100,
+      "description": "A collaborative art-and-technology experiment by Michelle Tylicki and Lauri Love that turns anti-surveillance makeup into an interactive teaching and testing experience.",
+      "visual_hint": "salon_mask",
+      "demonstrated": "The project demonstrates a salon-like interface for teaching Computer Vision Dazzle and testing camouflage against multiple facial recognition systems in an activist and public-learning context.",
+      "ghostati_relevance": "DAZZLE is one of Ghostati’s closest contemporary relatives: it combines makeup, interface design, quantitative feedback, and anti-surveillance pedagogy.",
+      "ghostati_testability": {
+        "level": "direct",
+        "reason": "Ghostati can directly retest the browser-lab logic of applying face-region visual interventions and comparing detection or matching behavior before and after the intervention."
+      },
+      "intervention": [
+        "makeup",
+        "face-obfuscation",
+        "interactive-tool",
+        "workshop"
+      ],
+      "target": [
+        "face-detection",
+        "face-recognition"
+      ],
+      "domain": [
+        "physical-world",
+        "browser",
+        "visible-light",
+        "festival"
+      ],
+      "limitations": "Public documentation describes the project and its testing ambition, but does not provide a full reproducible benchmark, exact model list, or protocol. Claims should be read as project documentation unless the implementation and results are independently released.",
+      "reproducibility": "documented-project",
+      "citation": "Tylicki and Love, DAZZLE, 2023",
+      "links": {
+        "canonical": "https://www.michelletylicki.info/dazzle/",
+        "paper": null,
+        "project": "https://www.michelletylicki.info/dazzle/",
+        "code": null,
+        "video": null,
+        "doi": null,
+        "arxiv": null
+      }
+    },
+    {
+      "slug": "physical-world-optical-attacks-3d-face",
+      "title": "Physical-World Optical Adversarial Attacks on 3D Face Recognition",
+      "author": [
+        "Yanjie Li",
+        "Yiquan Li",
+        "Xuelong Dai",
+        "Songtao Guo",
+        "Bin Xiao"
+      ],
+      "year": 2023,
+      "link": "https://openaccess.thecvf.com/content/CVPR2023/html/Li_Physical-World_Optical_Adversarial_Attacks_on_3D_Face_Recognition_CVPR_2023_paper.html",
+      "type": "research",
+      "closeness": 30,
+      "description": "A CVPR 2023 paper on structured-light optical attacks against 3D face recognition systems, moving anti-recognition intervention from visible surface patterns to depth reconstruction and projected perturbations.",
+      "visual_hint": "projector_depth_face",
+      "demonstrated": "The paper demonstrates a structured-light attack that can affect point-cloud-based and depth-image-based 3D face recognition with fewer perturbations and resilience to random head movements.",
+      "ghostati_relevance": "It helps Ghostati explain that face recognition is not only 2D RGB matching: depth, structured light, and skin reflectance introduce a different attack surface.",
+      "ghostati_testability": {
+        "level": "indirect",
+        "reason": "Ghostati’s browser pipeline can discuss optical perturbations conceptually, but cannot reproduce structured-light 3D scanning, depth-image recognition, or point-cloud attacks."
+      },
+      "intervention": [
+        "structured-light",
+        "optical",
+        "projection"
+      ],
+      "target": [
+        "3d-face-recognition",
+        "depth-face-recognition"
+      ],
+      "domain": [
+        "physical-world",
+        "optical",
+        "3d"
+      ],
+      "limitations": "The setup requires structured-light 3D recognition assumptions and specialized acquisition. It is not evidence that ordinary visible AR makeup affects 3D systems in the same way.",
+      "reproducibility": "paper-available",
+      "citation": "Li et al., CVPR 2023",
+      "links": {
+        "canonical": "https://openaccess.thecvf.com/content/CVPR2023/html/Li_Physical-World_Optical_Adversarial_Attacks_on_3D_Face_Recognition_CVPR_2023_paper.html",
+        "paper": "https://openaccess.thecvf.com/content/CVPR2023/papers/Li_Physical-World_Optical_Adversarial_Attacks_on_3D_Face_Recognition_CVPR_2023_paper.pdf",
+        "project": null,
+        "code": "https://github.com/PolyLiYJ/SLAttack",
+        "video": null,
+        "doi": null,
+        "arxiv": "https://arxiv.org/abs/2205.13412"
+      }
+    },
+    {
+      "slug": "camera-shy-hoodie",
+      "title": "The Camera-Shy Hoodie",
+      "author": [
+        "Mac Pierce"
+      ],
+      "year": 2023,
+      "link": "https://www.macpierce.com/the-camera-shy-hoodie",
+      "type": "artistic",
+      "closeness": 20,
+      "description": "A DIY wearable that uses high-power infrared LEDs to degrade or overexpose night-vision security camera footage around the wearer’s head and upper body.",
+      "visual_hint": "hoodie_ir",
+      "demonstrated": "The project demonstrates a buildable garment that interferes with IR-sensitive security cameras by using the same infrared wavelength commonly used for night-vision illumination.",
+      "ghostati_relevance": "It is a useful boundary reference: not face makeup and not face recognition in the strict sense, but a clear example of physical sensor disturbance as anti-surveillance design.",
+      "ghostati_testability": {
+        "level": "no",
+        "reason": "Ghostati cannot test IR LED glare, camera auto-exposure interference, or night-vision capture in a standard browser AR pipeline."
+      },
+      "intervention": [
+        "infrared",
+        "garment",
+        "sensor-jamming"
+      ],
+      "target": [
+        "security-camera",
+        "sensor-disturbance"
+      ],
+      "domain": [
+        "physical-world",
+        "near-infrared",
+        "wearable"
+      ],
+      "limitations": "Its effect depends on camera sensitivity, exposure behavior, LED placement, distance, and lighting. It does not directly test face-recognition matching or browser-based face analysis.",
+      "reproducibility": "build-files-available",
+      "citation": "Mac Pierce, The Camera-Shy Hoodie, 2023",
+      "links": {
+        "canonical": "https://www.macpierce.com/the-camera-shy-hoodie",
+        "paper": null,
+        "project": "https://www.macpierce.com/the-camera-shy-hoodie",
+        "code": null,
+        "video": null,
+        "doi": null,
+        "arxiv": null
+      }
+    },
+    {
+      "slug": "shadows-can-be-dangerous",
+      "title": "Shadows can be Dangerous: Stealthy and Effective Physical-world Adversarial Attack by Natural Phenomenon",
+      "author": [
+        "Yiqi Zhong",
+        "Xianming Liu",
+        "Deming Zhai",
+        "Junjun Jiang",
+        "Xiangyang Ji"
+      ],
+      "year": 2022,
+      "link": "https://arxiv.org/abs/2203.03818",
+      "type": "research",
+      "closeness": 35,
+      "description": "A physical-world adversarial attack that uses natural-looking shadows rather than stickers or conspicuous projected patterns, tested on traffic-sign recognition rather than face recognition.",
+      "visual_hint": "shadow_cast",
+      "demonstrated": "The paper demonstrates that optimized shadows can create stealthy black-box physical adversarial examples, with strong simulated and real-world attack results on traffic-sign datasets.",
+      "ghostati_relevance": "It is useful to Ghostati as a design reference for natural phenomena, lighting, and environmental perturbations, but it is not face-specific.",
+      "ghostati_testability": {
+        "level": "contextual",
+        "reason": "Ghostati can use this as a conceptual reference for light and shadow as adversarial surfaces, but cannot validate the traffic-sign attack or transfer it to face recognition without a new experiment."
+      },
+      "intervention": [
+        "shadow",
+        "optical",
+        "environmental-perturbation"
+      ],
+      "target": [
+        "traffic-sign-recognition",
+        "image-classification"
+      ],
+      "domain": [
+        "physical-world",
+        "visible-light",
+        "environmental"
+      ],
+      "limitations": "The demonstrated target is traffic-sign recognition, not face detection or recognition. Any relevance to Ghostati requires cautious analogy, not direct transfer of the result.",
+      "reproducibility": "code-available",
+      "citation": "Zhong et al., CVPR 2022",
+      "links": {
+        "canonical": "https://arxiv.org/abs/2203.03818",
+        "paper": "https://arxiv.org/abs/2203.03818",
+        "project": null,
+        "code": "https://github.com/hncszyq/ShadowAttack",
+        "video": null,
+        "doi": "https://doi.org/10.48550/arXiv.2203.03818",
+        "arxiv": "https://arxiv.org/abs/2203.03818"
+      }
+    },
+    {
+      "slug": "adv-makeup",
+      "title": "Adv-Makeup: A New Imperceptible and Transferable Attack on Face Recognition",
+      "author": [
+        "Bangjie Yin",
+        "Wenxuan Wang",
+        "Taiping Yao",
+        "Junfeng Guo",
+        "Zelun Kong",
+        "Shouhong Ding",
+        "Jilin Li",
+        "Cong Liu"
+      ],
+      "year": 2021,
+      "link": "https://arxiv.org/abs/2105.03162",
+      "type": "research",
+      "closeness": 95,
+      "description": "A key adversarial-makeup paper that synthesizes transferable, imperceptible eye-shadow perturbations over the orbital region to attack face recognition.",
+      "visual_hint": "orbital_makeup",
+      "demonstrated": "The paper demonstrates a task-driven makeup generation method and fine-grained meta-learning strategy intended to improve black-box transferability, including digital and physical scenarios.",
+      "ghostati_relevance": "This is one of the closest research references for Ghostati because it treats makeup as an adversarial surface on the machine-readable face.",
+      "ghostati_testability": {
+        "level": "partial",
+        "reason": "Ghostati can approximate orbital-region overlays and compare local recognition behavior, but it does not implement the paper’s adversarial optimization, meta-learning, or commercial-system evaluation protocol."
+      },
+      "intervention": [
+        "makeup",
+        "orbital-region",
+        "adversarial-optimization"
+      ],
+      "target": [
+        "face-recognition",
+        "face-verification"
+      ],
+      "domain": [
+        "physical-world",
+        "digital",
+        "visible-light"
+      ],
+      "limitations": "The reported performance depends on the models, datasets, optimization method, physical-printing or application process, and commercial APIs tested in the paper. A manually designed Ghostyle is not equivalent to Adv-Makeup optimization.",
+      "reproducibility": "paper-available",
+      "citation": "Yin et al., IJCAI 2021",
+      "links": {
+        "canonical": "https://arxiv.org/abs/2105.03162",
+        "paper": "https://arxiv.org/abs/2105.03162",
+        "project": null,
+        "code": null,
+        "video": null,
+        "doi": "https://doi.org/10.48550/arXiv.2105.03162",
+        "arxiv": "https://arxiv.org/abs/2105.03162"
+      }
+    },
+    {
+      "slug": "adversarial-attacks-against-face-recognition",
+      "title": "Adversarial Attacks against Face Recognition: A Comprehensive Study",
+      "author": [
+        "Fatemeh Vakhshiteh",
+        "Ahmad Nickabadi",
+        "Raghavendra Ramachandra"
+      ],
+      "year": 2020,
+      "link": "https://arxiv.org/abs/2007.11709",
+      "type": "research",
+      "closeness": 82,
+      "description": "A broad survey of adversarial attacks and defenses against face recognition systems, useful for situating Ghostati within a larger technical taxonomy.",
+      "visual_hint": "survey_grid",
+      "demonstrated": "The article surveys attack and defense methods, proposes taxonomies, compares criteria, and identifies challenges and future research directions for adversarial face recognition.",
+      "ghostati_relevance": "It gives Ghostati a technical map for classifying references by threat model, physical/digital domain, target stage, and defense assumptions.",
+      "ghostati_testability": {
+        "level": "contextual",
+        "reason": "Ghostati can use the survey to classify experiments and limitations, but a survey is not itself a system or attack to retest."
+      },
+      "intervention": [
+        "survey",
+        "taxonomy"
+      ],
+      "target": [
+        "face-recognition",
+        "face-verification"
+      ],
+      "domain": [
+        "digital",
+        "physical-world"
+      ],
+      "limitations": "As a survey, it summarizes prior work rather than providing one reproducible experimental setup. Some included systems and defenses may have changed since publication.",
+      "reproducibility": "survey",
+      "citation": "Vakhshiteh, Nickabadi, and Ramachandra, 2020/2021",
+      "links": {
+        "canonical": "https://arxiv.org/abs/2007.11709",
+        "paper": "https://arxiv.org/abs/2007.11709",
+        "project": null,
+        "code": null,
+        "video": null,
+        "doi": "https://doi.org/10.48550/arXiv.2007.11709",
+        "arxiv": "https://arxiv.org/abs/2007.11709"
+      }
+    },
+    {
+      "slug": "breaking-certified-defenses-shadow-attack",
+      "title": "Breaking certified defenses: Semantic adversarial examples with spoofed robustness certificates",
+      "author": [
+        "Amin Ghiasi",
+        "Ali Shafahi",
+        "Tom Goldstein"
+      ],
+      "year": 2020,
+      "link": "https://arxiv.org/abs/2003.08937",
+      "type": "research",
+      "closeness": 35,
+      "description": "A semantic adversarial-example paper that uses large but perceptually plausible perturbations to attack robust classifiers and spoof robustness certificates.",
+      "visual_hint": "shadow_cast",
+      "demonstrated": "The paper demonstrates a Shadow Attack that can cause certified classifiers to mislabel images while still producing a spoofed robustness certificate.",
+      "ghostati_relevance": "It is conceptually relevant because it frames adversarial design as a semantic, human-plausible transformation rather than only pixel noise, but it is not face- or makeup-specific.",
+      "ghostati_testability": {
+        "level": "contextual",
+        "reason": "Ghostati can borrow the idea of perceptually plausible visual transformations, but cannot retest certified robustness or the paper’s classifier/certificate setting."
+      },
+      "intervention": [
+        "shadow",
+        "semantic-perturbation"
+      ],
+      "target": [
+        "image-classification",
+        "certified-defense"
+      ],
+      "domain": [
+        "digital"
+      ],
+      "limitations": "The work is not face-specific and addresses certified image classifiers rather than browser face matching. Relevance to Ghostati is conceptual and should not be presented as evidence for face-camouflage effectiveness.",
+      "reproducibility": "paper-available",
+      "citation": "Ghiasi, Shafahi, and Goldstein, 2020",
+      "links": {
+        "canonical": "https://arxiv.org/abs/2003.08937",
+        "paper": "https://arxiv.org/abs/2003.08937",
+        "project": null,
+        "code": null,
+        "video": null,
+        "doi": "https://doi.org/10.48550/arXiv.2003.08937",
+        "arxiv": "https://arxiv.org/abs/2003.08937"
+      }
+    },
+    {
+      "slug": "dazzle-club",
+      "title": "The Dazzle Club",
+      "author": [
+        "Evie Price",
+        "Emily Roderick",
+        "Georgina Rowlands",
+        "Anna Hart"
+      ],
+      "year": 2019,
+      "link": "https://emilyroderick.com/work/the-dazzle-club/",
+      "type": "activism",
+      "closeness": 96,
+      "description": "A London-based collective practice using CV Dazzle walks, workshops, films, and public-space research to draw attention to facial recognition and surveillance infrastructure.",
+      "visual_hint": "group_walk",
+      "period": "2019–2021",
+      "demonstrated": "The Dazzle Club demonstrates how face paint and CV Dazzle can operate as embodied public research, collective performance, and public education rather than only as a technical evasion method.",
+      "ghostati_relevance": "It helps Ghostati understand camouflage as a social and workshop practice: a way to make public surveillance visible, discussable, and collectively testable.",
+      "ghostati_testability": {
+        "level": "contextual",
+        "reason": "Ghostati can support similar workshops and local tests, but it cannot reproduce the public-space performance, legal context, or collective embodied research of the project."
+      },
+      "intervention": [
+        "makeup",
+        "performance",
+        "public-walk",
+        "workshop"
+      ],
+      "target": [
+        "public-space-surveillance",
+        "face-recognition"
+      ],
+      "domain": [
+        "physical-world",
+        "public-space",
+        "festival"
+      ],
+      "limitations": "The project is primarily an artistic and activist practice, not a controlled technical benchmark. Its value is public research and awareness rather than a reproducible recognition-evasion result.",
+      "reproducibility": "artwork-documentation",
+      "citation": "The Dazzle Club, 2019–2021",
+      "links": {
+        "canonical": "https://emilyroderick.com/work/the-dazzle-club/",
+        "paper": null,
+        "project": "https://emilyroderick.com/work/the-dazzle-club/",
+        "code": null,
+        "video": null,
+        "doi": null,
+        "arxiv": null
+      }
+    },
+    {
+      "slug": "vla-visible-light-face-attack",
+      "title": "VLA: A Practical Visible Light-based Attack on Face Recognition Systems in Physical World",
+      "author": [
+        "Meng Shen",
+        "Zelin Liao",
+        "Liehuang Zhu",
+        "Ke Xu",
+        "Xiaojiang Du"
+      ],
+      "year": 2019,
+      "link": "https://doi.org/10.1145/3351261",
+      "type": "research",
+      "closeness": 78,
+      "description": "A visible-light physical attack that projects adversarial perturbations onto a face to conduct targeted or untargeted attacks against black-box face recognition systems.",
+      "visual_hint": "light_beam",
+      "demonstrated": "The paper proposes VLA, a visible-light projection method for crafting physical adversarial perturbations on the face against black-box face recognition systems.",
+      "ghostati_relevance": "It connects Ghostati’s visual overlay approach to optical projection attacks: both change the appearance captured by the camera, but VLA is optimized and physically projected rather than drawn as a browser layer.",
+      "ghostati_testability": {
+        "level": "partial",
+        "reason": "Ghostati can simulate a visible overlay on the face and measure local matching changes, but it cannot validate projector optics, camera-image formation, or the paper’s black-box attack setup."
+      },
+      "intervention": [
+        "light",
+        "projection",
+        "optical"
+      ],
+      "target": [
+        "face-recognition",
+        "face-verification"
+      ],
+      "domain": [
+        "physical-world",
+        "visible-light",
+        "optical"
+      ],
+      "limitations": "The result depends on projector/camera geometry, brightness, environmental lighting, attack optimization, and the target recognition systems. Browser AR overlays are not equivalent to physically projected light.",
+      "reproducibility": "paper-available",
+      "citation": "Shen et al., UbiComp/IMWUT 2019",
+      "links": {
+        "canonical": "https://doi.org/10.1145/3351261",
+        "paper": "https://www.thucsnet.com/wp-content/papers/meng_ubicomp19.pdf",
+        "project": null,
+        "code": null,
+        "video": null,
+        "doi": "https://doi.org/10.1145/3351261",
+        "arxiv": null
+      }
+    },
+    {
+      "slug": "art-toolbox",
+      "title": "Adversarial Robustness Toolbox v1.0.0",
+      "author": [
+        "Maria-Irina Nicolae",
+        "Mathieu Sinn",
+        "Minh Ngoc Tran",
+        "Beat Buesser",
+        "Ambrish Rawat",
+        "Martin Wistuba",
+        "Valentina Zantedeschi",
+        "Nathalie Baracaldo",
+        "Bryant Chen",
+        "Heiko Ludwig",
+        "Ian M. Molloy",
+        "Ben Edwards"
+      ],
+      "year": 2018,
+      "link": "https://arxiv.org/abs/1807.01069",
+      "type": "research",
+      "closeness": 55,
+      "description": "A general-purpose Python library for testing machine-learning models with adversarial attacks and defenses, useful as technical infrastructure rather than as a face-camouflage reference.",
+      "visual_hint": "toolbox",
+      "demonstrated": "ART provides implementations, examples, and documentation for attacks and defenses across many model families and machine-learning frameworks.",
+      "ghostati_relevance": "It helps situate Ghostati inside adversarial ML practice, but it is not a visual face-camouflage project by itself.",
+      "ghostati_testability": {
+        "level": "indirect",
+        "reason": "Ghostati could conceptually integrate adversarial-evaluation workflows, but ART does not directly map to the current browser-only Ghostyle test loop."
+      },
+      "intervention": [
+        "toolkit",
+        "adversarial-ml"
+      ],
+      "target": [
+        "machine-learning-models"
+      ],
+      "domain": [
+        "digital",
+        "software-library"
+      ],
+      "limitations": "ART is broad infrastructure, not a face-specific or physical-world camouflage result. Applying it to Ghostati would require additional model wrappers and experiment design.",
+      "reproducibility": "code-available",
+      "citation": "Nicolae et al., 2018/2019",
+      "links": {
+        "canonical": "https://arxiv.org/abs/1807.01069",
+        "paper": "https://arxiv.org/abs/1807.01069",
+        "project": null,
+        "code": "https://github.com/Trusted-AI/adversarial-robustness-toolbox",
+        "video": null,
+        "doi": "https://doi.org/10.48550/arXiv.1807.01069",
+        "arxiv": "https://arxiv.org/abs/1807.01069"
+      }
+    },
+    {
+      "slug": "dpatch",
+      "title": "DPatch: An Adversarial Patch Attack on Object Detectors",
+      "author": [
+        "Xin Liu",
+        "Huanrui Yang",
+        "Ziwei Liu",
+        "Linghao Song",
+        "Hai Li",
+        "Yiran Chen"
+      ],
+      "year": 2018,
+      "link": "https://arxiv.org/abs/1806.02299",
+      "type": "research",
+      "closeness": 40,
+      "description": "A black-box adversarial patch attack against object detectors such as Faster R-CNN and YOLO, relevant to the broader genealogy of physical adversarial patches.",
+      "visual_hint": "detector_box_patch",
+      "demonstrated": "DPatch demonstrates that a small, location-independent adversarial patch can attack both bounding-box regression and object classification in object detectors, with transferability across detectors.",
+      "ghostati_relevance": "It is not face-specific, but it explains why localized graphic surfaces can disrupt detection pipelines, a core idea behind many physical-world adversarial designs.",
+      "ghostati_testability": {
+        "level": "indirect",
+        "reason": "Ghostati can draw face-localized patches, but DPatch targets object detectors and requires detector-specific optimization that Ghostati does not implement."
+      },
+      "intervention": [
+        "patch",
+        "adversarial-optimization"
+      ],
+      "target": [
+        "object-detection"
+      ],
+      "domain": [
+        "digital",
+        "physical-world",
+        "printed"
+      ],
+      "limitations": "The attack target is object detection, not face recognition. Transfer to face analysis is conceptual unless a new face-specific optimization and evaluation are performed.",
+      "reproducibility": "paper-available",
+      "citation": "Liu et al., SafeAI 2019 / arXiv 2018",
+      "links": {
+        "canonical": "https://arxiv.org/abs/1806.02299",
+        "paper": "https://arxiv.org/abs/1806.02299",
+        "project": null,
+        "code": null,
+        "video": null,
+        "doi": "https://doi.org/10.48550/arXiv.1806.02299",
+        "arxiv": "https://arxiv.org/abs/1806.02299"
+      }
+    },
+    {
+      "slug": "shapeshifter",
+      "title": "ShapeShifter: Robust Physical Adversarial Attack on Faster R-CNN Object Detector",
+      "author": [
+        "Shang-Tse Chen",
+        "Cory Cornelius",
+        "Jason Martin",
+        "Duen Horng Chau"
+      ],
+      "year": 2018,
+      "link": "https://arxiv.org/abs/1804.05810",
+      "type": "research",
+      "closeness": 38,
+      "description": "A physical adversarial attack on Faster R-CNN object detection that adapts robust transformation methods to real-world detector attacks.",
+      "visual_hint": "object_shape_distort",
+      "demonstrated": "ShapeShifter shows that physical perturbations can be optimized to survive real-world transformations such as distance, angle, lighting, and camera limitations in an object-detection setting.",
+      "ghostati_relevance": "It is a foundational physical-world adversarial reference: not face-centered, but important for understanding robustness under real-world transformations.",
+      "ghostati_testability": {
+        "level": "contextual",
+        "reason": "Ghostati can borrow its concern with transformations, but cannot directly retest Faster R-CNN stop-sign attacks or autonomous-driving scenarios."
+      },
+      "intervention": [
+        "patch",
+        "printed",
+        "adversarial-optimization"
+      ],
+      "target": [
+        "object-detection"
+      ],
+      "domain": [
+        "physical-world",
+        "visible-light",
+        "printed"
+      ],
+      "limitations": "The target domain is object detection, not biometric recognition. Its lessons transfer as methodology, not as a direct Ghostati result.",
+      "reproducibility": "paper-available",
+      "citation": "Chen et al., ECML PKDD 2018",
+      "links": {
+        "canonical": "https://arxiv.org/abs/1804.05810",
+        "paper": "https://arxiv.org/abs/1804.05810",
+        "project": null,
+        "code": null,
+        "video": null,
+        "doi": "https://doi.org/10.1007/978-3-030-10925-7_4",
+        "arxiv": "https://arxiv.org/abs/1804.05810"
+      }
+    },
+    {
+      "slug": "agns-face-recognition",
+      "title": "A General Framework for Adversarial Examples with Objectives",
+      "author": [
+        "Mahmood Sharif",
+        "Sruti Bhagavatula",
+        "Lujo Bauer",
+        "Michael K. Reiter"
+      ],
+      "year": 2017,
+      "link": "https://arxiv.org/abs/1801.00349",
+      "type": "research",
+      "closeness": 90,
+      "description": "The later framework paper behind adversarial generative nets, including physical eyeglass frames designed to fool face recognition with better robustness, inconspicuousness, and scalability than earlier approaches.",
+      "visual_hint": "generated_glasses",
+      "demonstrated": "The paper demonstrates adversarial generative nets that can satisfy application-specific objectives, including physical eyeglass-frame examples for face recognition.",
+      "ghostati_relevance": "It is highly relevant because it treats facial accessories as optimized physical adversarial media, adjacent to Ghostati’s Ghostyle concept even though Ghostati uses browser-rendered visual styles rather than generated eyeglass frames.",
+      "ghostati_testability": {
+        "level": "partial",
+        "reason": "Ghostati can retest the broad idea of face-region or accessory-like overlays, but not the generator training, eyeglass printing pipeline, or exact attack objective."
+      },
+      "intervention": [
+        "glasses",
+        "wearable-accessory",
+        "adversarial-optimization"
+      ],
+      "target": [
+        "face-recognition",
+        "face-verification"
+      ],
+      "domain": [
+        "physical-world",
+        "visible-light",
+        "wearable"
+      ],
+      "limitations": "The result depends on generated eyeglass-frame patterns, model assumptions, and physical realization. Browser overlays can approximate placement but not validate the original optimization or robustness claims.",
+      "reproducibility": "paper-available",
+      "citation": "Sharif et al., arXiv 2017 / ACM TOPS 2019",
+      "links": {
+        "canonical": "https://arxiv.org/abs/1801.00349",
+        "paper": "https://arxiv.org/abs/1801.00349",
+        "project": null,
+        "code": null,
+        "video": null,
+        "doi": "https://doi.org/10.1145/3317611",
+        "arxiv": "https://arxiv.org/abs/1801.00349"
+      }
+    },
+    {
+      "slug": "adversarial-patch",
+      "title": "Adversarial Patch",
+      "author": [
+        "Tom B. Brown",
+        "Dandelion Mané",
+        "Aurko Roy",
+        "Martín Abadi",
+        "Justin Gilmer"
+      ],
+      "year": 2017,
+      "link": "https://arxiv.org/abs/1712.09665",
+      "type": "research",
+      "closeness": 50,
+      "description": "A seminal paper on universal, robust, targeted adversarial image patches that can be printed, added to scenes, photographed, and presented to classifiers.",
+      "visual_hint": "patch_sticker",
+      "demonstrated": "The work demonstrates printable universal patches that can cause classifiers to ignore other scene content and output a chosen target class across transformations.",
+      "ghostati_relevance": "It provides the core technical idea behind many visible adversarial surfaces: a localized graphic object can dominate a model’s interpretation of an image.",
+      "ghostati_testability": {
+        "level": "indirect",
+        "reason": "Ghostati can draw patch-like overlays, but the original paper targets image classifiers and requires optimization not present in Ghostati."
+      },
+      "intervention": [
+        "patch",
+        "printed",
+        "adversarial-optimization"
+      ],
+      "target": [
+        "image-classification"
+      ],
+      "domain": [
+        "physical-world",
+        "printed",
+        "visible-light"
+      ],
+      "limitations": "The paper is not face-specific. Its relevance to Ghostati is methodological unless a face-recognition-specific patch is optimized and tested.",
+      "reproducibility": "code-available",
+      "citation": "Brown et al., 2017/2018",
+      "links": {
+        "canonical": "https://arxiv.org/abs/1712.09665",
+        "paper": "https://arxiv.org/abs/1712.09665",
+        "project": null,
+        "code": "https://github.com/tensorflow/cleverhans/tree/master/examples/adversarial_patch",
+        "video": null,
+        "doi": "https://doi.org/10.48550/arXiv.1712.09665",
+        "arxiv": "https://arxiv.org/abs/1712.09665"
+      }
+    },
+    {
+      "slug": "accessorize-to-a-crime",
+      "title": "Accessorize to a Crime: Real and Stealthy Attacks on State-of-the-Art Face Recognition",
+      "author": [
+        "Mahmood Sharif",
+        "Sruti Bhagavatula",
+        "Lujo Bauer",
+        "Michael K. Reiter"
+      ],
+      "year": 2016,
+      "link": "https://dl.acm.org/doi/10.1145/2976749.2978392",
+      "type": "research",
+      "closeness": 92,
+      "description": "A landmark physical attack on face recognition using printed eyeglass frames to support evasion or impersonation attacks in a comparatively inconspicuous form.",
+      "visual_hint": "glasses",
+      "demonstrated": "The paper defines physically realizable and inconspicuous attacks on facial biometric systems and realizes them by printing adversarial eyeglass frames.",
+      "ghostati_relevance": "It is one of the strongest technical ancestors for Ghostati’s appearance-based tests: the face is not hidden, but visually modified to change machine matching.",
+      "ghostati_testability": {
+        "level": "partial",
+        "reason": "Ghostati can approximate glasses-like overlays and compare local matching behavior, but cannot reproduce printed-material effects, physical-camera capture, or the exact optimization pipeline."
+      },
+      "intervention": [
+        "glasses",
+        "wearable-accessory",
+        "adversarial-optimization"
+      ],
+      "target": [
+        "face-recognition",
+        "face-verification"
+      ],
+      "domain": [
+        "physical-world",
+        "visible-light",
+        "wearable"
+      ],
+      "limitations": "The results are tied to the target systems, printed patterns, imaging conditions, and experimental subjects. Newer recognition models or production systems require fresh testing.",
+      "reproducibility": "code-available",
+      "citation": "Sharif et al., ACM CCS 2016",
+      "links": {
+        "canonical": "https://doi.org/10.1145/2976749.2978392",
+        "paper": "https://users.ece.cmu.edu/~lbauer/papers/2016/ccs2016-face-recognition.pdf",
+        "project": "https://mahmoods01.github.io/publication/ccs16-adv-ml/",
+        "code": "https://github.com/mahmoods01/accessorize-to-a-crime",
+        "video": null,
+        "doi": "https://doi.org/10.1145/2976749.2978392",
+        "arxiv": null
+      }
+    },
+    {
+      "slug": "hyperface",
+      "title": "HyperFace",
+      "author": [
+        "Adam Harvey"
+      ],
+      "year": 2016,
+      "link": "https://adam.harvey.studio/hyperface/",
+      "type": "artistic",
+      "closeness": 92,
+      "description": "Adam Harvey’s false-face textile camouflage, developed with Hyphen-Labs, that attempts to redirect face-detection confidence toward surrounding decoy faces rather than the wearer’s real face.",
+      "visual_hint": "false_faces",
+      "demonstrated": "HyperFace demonstrates a ground-based camouflage strategy: instead of hiding the face directly, it supplies machine-readable false faces around it to distract specific detectors.",
+      "ghostati_relevance": "It expands Ghostati’s design vocabulary from face overlays to surrounding false-face patterns and attention redirection.",
+      "ghostati_testability": {
+        "level": "partial",
+        "reason": "Ghostati can test surrounding decoy-face graphics against browser detection in a limited way, but the original HyperFace patterns targeted specific Haarcascade detectors and textile contexts."
+      },
+      "intervention": [
+        "textile",
+        "false-face",
+        "decoy",
+        "pattern"
+      ],
+      "target": [
+        "face-detection"
+      ],
+      "domain": [
+        "physical-world",
+        "visible-light",
+        "wearable"
+      ],
+      "limitations": "The documented prototype targeted OpenCV Haarcascade-style detectors and is not designed or tested against modern deep CNN-based face detection systems. Its value as camouflage is explicitly temporary and system-specific.",
+      "reproducibility": "documented-project",
+      "citation": "Adam Harvey, HyperFace, 2016/2017",
+      "links": {
+        "canonical": "https://adam.harvey.studio/hyperface/",
+        "paper": null,
+        "project": "https://adam.harvey.studio/hyperface/",
+        "code": null,
+        "video": null,
+        "doi": null,
+        "arxiv": null
+      }
+    },
+    {
+      "slug": "adversarial-manipulation-deep-representations",
+      "title": "Adversarial Manipulation of Deep Representations",
+      "author": [
+        "Sara Sabour",
+        "Yanshuai Cao",
+        "Fartash Faghri",
+        "David J. Fleet"
+      ],
+      "year": 2015,
+      "link": "https://arxiv.org/abs/1511.05122",
+      "type": "research",
+      "closeness": 45,
+      "description": "A theoretical reference showing that an image’s internal deep-network representation can be manipulated to resemble another image while remaining visually similar to the original.",
+      "visual_hint": "latent_match",
+      "demonstrated": "The paper demonstrates feature adversaries: small perturbations that make internal DNN representations mimic different natural images, raising questions about what representations encode.",
+      "ghostati_relevance": "It gives Ghostati a conceptual basis for thinking beyond labels: recognition systems operate through internal representations, not human-readable identity alone.",
+      "ghostati_testability": {
+        "level": "contextual",
+        "reason": "Ghostati can discuss representation shifts, but it does not expose or optimize internal DNN layers in the current browser pipeline."
+      },
+      "intervention": [
+        "digital-perturbation",
+        "representation-attack"
+      ],
+      "target": [
+        "deep-representation",
+        "image-classification"
+      ],
+      "domain": [
+        "digital"
+      ],
+      "limitations": "This is not a physical-world or face-specific intervention. Its connection to Ghostati is conceptual unless the project adds model-internal representation testing.",
+      "reproducibility": "paper-available",
+      "citation": "Sabour et al., ICLR 2016",
+      "links": {
+        "canonical": "https://arxiv.org/abs/1511.05122",
+        "paper": "https://arxiv.org/abs/1511.05122",
+        "project": null,
+        "code": null,
+        "video": null,
+        "doi": "https://doi.org/10.48550/arXiv.1511.05122",
+        "arxiv": "https://arxiv.org/abs/1511.05122"
+      }
+    },
+    {
+      "slug": "cv-dazzle",
+      "title": "CV Dazzle",
+      "author": [
+        "Adam Harvey"
+      ],
+      "year": 2010,
+      "link": "https://adam.harvey.studio/cvdazzle/",
+      "type": "artistic",
+      "closeness": 100,
+      "description": "Adam Harvey’s foundational computer-vision camouflage project, created in 2010, using makeup, hairstyling, and asymmetry to disrupt face detection while remaining visible to humans.",
+      "visual_hint": "dazzle_face",
+      "demonstrated": "CV Dazzle demonstrated that hair and makeup alone could lower detection probability below threshold for several OpenCV Haarcascade face-detection profiles in the original proof-of-concept setting.",
+      "ghostati_relevance": "It is the central historical anchor for Ghostati: a technique-first, system-specific approach to anti-biometric appearance design.",
+      "ghostati_testability": {
+        "level": "direct",
+        "reason": "Ghostati can retest the general strategy of visible face-region interventions against its own browser detection and matching pipeline, while clearly separating old Haarcascade claims from modern models."
+      },
+      "intervention": [
+        "makeup",
+        "hair",
+        "asymmetry",
+        "face-obfuscation"
+      ],
+      "target": [
+        "face-detection"
+      ],
+      "domain": [
+        "physical-world",
+        "visible-light"
+      ],
+      "limitations": "The original looks were designed for Viola-Jones/OpenCV Haarcascade detectors and should not be treated as active protection against modern CNN-based detection or recognition systems. Lighting, pose, algorithm, and wearer-specific design all matter.",
+      "reproducibility": "documented-project",
+      "citation": "Adam Harvey, CV Dazzle, 2010",
+      "links": {
+        "canonical": "https://adam.harvey.studio/cvdazzle/",
+        "paper": null,
+        "project": "https://adam.harvey.studio/cvdazzle/",
+        "code": null,
+        "video": null,
+        "doi": null,
+        "arxiv": null
+      }
+    }
+  ]
+}
diff --git a/references/REFERENCES.schema.json b/references/REFERENCES.schema.json
new file mode 100644
index 0000000..22af4a7
--- /dev/null
+++ b/references/REFERENCES.schema.json
@@ -0,0 +1,116 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://sindacato.nina.watch/ghostati/REFERENCES.schema.json",
+  "title": "Ghostati References",
+  "type": "object",
+  "required": [
+    "schema_version",
+    "project_scope",
+    "ordering_definition",
+    "language",
+    "closeness_definition",
+    "tag_definitions",
+    "references"
+  ],
+  "properties": {
+    "schema_version": { "type": "string", "pattern": "^2\\.[0-9]+$" },
+    "project_scope": { "type": "string", "minLength": 1 },
+    "ordering_definition": { "type": "string", "minLength": 1 },
+    "language": { "type": "string", "const": "en" },
+    "closeness_definition": { "type": "object" },
+    "tag_definitions": {
+      "type": "object",
+      "required": ["type", "ghostati_testability_level", "intervention", "target", "domain"],
+      "properties": {
+        "type": { "type": "array", "items": { "enum": ["artistic", "research", "activism"] }, "uniqueItems": true },
+        "ghostati_testability_level": { "type": "array", "items": { "enum": ["direct", "partial", "indirect", "contextual", "no"] }, "uniqueItems": true },
+        "intervention": { "type": "array", "items": { "type": "string", "pattern": "^[a-z0-9-]+$" }, "uniqueItems": true },
+        "target": { "type": "array", "items": { "type": "string", "pattern": "^[a-z0-9-]+$" }, "uniqueItems": true },
+        "domain": { "type": "array", "items": { "type": "string", "pattern": "^[a-z0-9-]+$" }, "uniqueItems": true }
+      },
+      "additionalProperties": false
+    },
+    "references": {
+      "type": "array",
+      "minItems": 1,
+      "items": { "$ref": "#/$defs/reference" }
+    }
+  },
+  "additionalProperties": false,
+  "$defs": {
+    "nullableUrl": {
+      "anyOf": [
+        { "type": "null" },
+        { "type": "string", "pattern": "^https?://" }
+      ]
+    },
+    "tag": { "type": "string", "pattern": "^[a-z0-9-]+$" },
+    "reference": {
+      "type": "object",
+      "required": [
+        "slug",
+        "title",
+        "author",
+        "year",
+        "link",
+        "links",
+        "type",
+        "closeness",
+        "description",
+        "demonstrated",
+        "ghostati_relevance",
+        "ghostati_testability",
+        "intervention",
+        "target",
+        "domain",
+        "limitations",
+        "reproducibility",
+        "citation"
+      ],
+      "properties": {
+        "slug": { "type": "string", "pattern": "^[a-z0-9]+(?:-[a-z0-9]+)*$" },
+        "title": { "type": "string", "minLength": 1 },
+        "author": { "type": "array", "minItems": 1, "items": { "type": "string", "minLength": 1 } },
+        "year": { "type": "integer", "minimum": 1900, "maximum": 2100 },
+        "period": { "type": "string" },
+        "link": { "type": "string", "pattern": "^https?://" },
+        "links": {
+          "type": "object",
+          "required": ["canonical", "paper", "project", "code", "video", "doi", "arxiv"],
+          "properties": {
+            "canonical": { "$ref": "#/$defs/nullableUrl" },
+            "paper": { "$ref": "#/$defs/nullableUrl" },
+            "project": { "$ref": "#/$defs/nullableUrl" },
+            "code": { "$ref": "#/$defs/nullableUrl" },
+            "video": { "$ref": "#/$defs/nullableUrl" },
+            "doi": { "$ref": "#/$defs/nullableUrl" },
+            "arxiv": { "$ref": "#/$defs/nullableUrl" }
+          },
+          "additionalProperties": true
+        },
+        "type": { "enum": ["artistic", "research", "activism"] },
+        "closeness": { "type": "integer", "minimum": 1, "maximum": 100 },
+        "description": { "type": "string", "minLength": 1 },
+        "demonstrated": { "type": "string", "minLength": 1 },
+        "ghostati_relevance": { "type": "string", "minLength": 1 },
+        "ghostati_testability": {
+          "type": "object",
+          "required": ["level", "reason"],
+          "properties": {
+            "level": { "enum": ["direct", "partial", "indirect", "contextual", "no"] },
+            "reason": { "type": "string", "minLength": 1 }
+          },
+          "additionalProperties": false
+        },
+        "intervention": { "type": "array", "minItems": 1, "items": { "$ref": "#/$defs/tag" }, "uniqueItems": true },
+        "target": { "type": "array", "minItems": 1, "items": { "$ref": "#/$defs/tag" }, "uniqueItems": true },
+        "domain": { "type": "array", "minItems": 1, "items": { "$ref": "#/$defs/tag" }, "uniqueItems": true },
+        "limitations": { "type": "string", "minLength": 1 },
+        "reproducibility": { "type": "string", "minLength": 1 },
+        "citation": { "type": "string", "minLength": 1 },
+        "visual_hint": { "type": "string" }
+      },
+      "additionalProperties": true
+    }
+  }
+}
diff --git a/references/build-references-page.js b/references/build-references-page.js
new file mode 100644
index 0000000..33383de
--- /dev/null
+++ b/references/build-references-page.js
@@ -0,0 +1,295 @@
+#!/usr/bin/env node
+'use strict';
+
+/**
+ * Regenerates references/index.html from REFERENCES.json + templates/references.template.html.
+ *
+ * Usage:
+ *   node references/validate-references.js && node references/build-references-page.js
+ *
+ * What's static vs. generated:
+ *   - templates/references.template.html holds the header, nav, intro copy, the
+ *     "How to suggest a new entry" section, and the aside shell. Edit that file
+ *     directly for wording/nav changes — this script never touches it, it only
+ *     fills in the {{...}} placeholders it contains (including the one-time
+ *     star icon sprite used for the relevance rating — swap the <path> there
+ *     whenever you have a real icon; every star instance updates automatically).
+ *   - Everything under {{REFERENCES_BODY}} is generated from REFERENCES.json
+ *     every time this runs, grouped by type: peer-reviewed papers, art
+ *     initiatives, advocacy efforts, then anything else.
+ *
+ * Two things are computed here rather than stored in the JSON, on purpose —
+ * the JSON schema hasn't changed:
+ *   - relevance stars (1-5): bucketed from the existing 1-100 "closeness"
+ *     score, using the same five bands already defined in
+ *     PROMPT-REFERENCES-UPDATE.txt's closeness rubric.
+ *   - "only here" / "first seen" tag flags: computed once across the whole
+ *     dataset (how many entries use a tag, and which entry used it earliest),
+ *     so a tag chip can point out when it's rare or when this is the entry
+ *     that introduced it, without needing clickable filtering.
+ *
+ * This script has no npm dependencies — plain Node + fs/path only.
+ */
+
+const fs = require('fs');
+const path = require('path');
+const { execFileSync } = require('child_process');
+
+const ROOT = path.resolve(__dirname, '..');
+const REFERENCES_PATH = path.join(ROOT, 'references', 'REFERENCES.json');
+const TEMPLATE_PATH = path.join(ROOT, 'references', 'templates', 'references.template.html');
+const OUTPUT_PATH = path.join(ROOT, 'references', 'index.html');
+const VALIDATOR_PATH = path.join(ROOT, 'references', 'validate-references.js');
+
+const TESTABILITY_LABEL = {
+  direct: 'Direct testability',
+  partial: 'Partial testability',
+  indirect: 'Indirect testability',
+  contextual: 'Contextual reference',
+  no: 'Not testable in-browser',
+};
+
+const LINK_LABELS = [
+  ['canonical', 'Canonical'],
+  ['paper', 'Paper'],
+  ['project', 'Project'],
+  ['code', 'Code'],
+  ['video', 'Video'],
+  ['doi', 'DOI'],
+  ['arxiv', 'arXiv'],
+];
+
+// Order matters: this is both the display order of the sections in the page
+// and the order of links in the "Jump to" nav.
+const GROUP_DEFS = [
+  { key: 'research', label: 'Peer-reviewed papers', anchor: 'group-research' },
+  { key: 'artistic', label: 'Art initiatives', anchor: 'group-artistic' },
+  { key: 'activism', label: 'Advocacy efforts', anchor: 'group-activism' },
+];
+// Anything whose "type" isn't one of the three above lands here. Empty today,
+// but this keeps the page from silently dropping an entry if a new type
+// value shows up later.
+const OTHER_GROUP = { key: 'other', label: 'Other', anchor: 'group-other' };
+
+function esc(value) {
+  return String(value)
+    .replace(/&/g, '&amp;')
+    .replace(/</g, '&lt;')
+    .replace(/>/g, '&gt;')
+    .replace(/"/g, '&quot;');
+}
+
+function capitalize(s) {
+  return s.charAt(0).toUpperCase() + s.slice(1);
+}
+
+function sortReferences(refs) {
+  // Matches REFERENCES.json's own ordering_definition:
+  // year desc, closeness desc, title asc.
+  return [...refs].sort((a, b) => {
+    if (a.year !== b.year) return b.year - a.year;
+    if (a.closeness !== b.closeness) return b.closeness - a.closeness;
+    return a.title.localeCompare(b.title);
+  });
+}
+
+// closeness -> 1-5 stars, using the exact bands from the closeness rubric:
+// 100 / 90-99 / 70-89 / 40-69 / 1-39. Five bands in, five stars out.
+function relevanceStars(closeness) {
+  if (closeness >= 100) return 5;
+  if (closeness >= 90) return 4;
+  if (closeness >= 70) return 3;
+  if (closeness >= 40) return 2;
+  return 1;
+}
+
+function starsHtml(closeness) {
+  const n = relevanceStars(closeness);
+  let icons = '';
+  for (let i = 1; i <= 5; i++) {
+    const empty = i > n;
+    icons += `<svg class="star-icon${empty ? ' star-icon--empty' : ''}" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg>`;
+  }
+  return `<span class="reference-relevance" title="Relevance ${n}/5 (source score ${esc(closeness)}/100)"><span class="sr-only">Relevance: ${n} out of 5</span>${icons}</span>`;
+}
+
+// One pass over the whole dataset: for every tag used in intervention/target/
+// domain, how many entries use it (deduped per entry) and the earliest year
+// it shows up in. Order-independent, so it doesn't matter that this runs
+// before sorting/grouping.
+function buildTagStats(refs) {
+  const stats = new Map(); // tag -> { count, minYear }
+  for (const ref of refs) {
+    const tags = new Set([...ref.intervention, ...ref.target, ...ref.domain]);
+    for (const tag of tags) {
+      const entry = stats.get(tag) || { count: 0, minYear: Infinity };
+      entry.count += 1;
+      entry.minYear = Math.min(entry.minYear, ref.year);
+      stats.set(tag, entry);
+    }
+  }
+  return stats;
+}
+
+function tagRow(ref) {
+  const level = ref.ghostati_testability.level;
+  const parts = [
+    `<span class="tag">${esc(ref.year)}</span>`,
+    `<span class="tag">${esc(capitalize(ref.type))}</span>`,
+    starsHtml(ref.closeness),
+    `<span class="badge badge--${esc(level)}">${esc(TESTABILITY_LABEL[level] || level)}</span>`,
+  ];
+  return `<div class="reference-entry__meta">${parts.join('')}</div>`;
+}
+
+function topicTags(ref, tagStats) {
+  const tags = [...new Set([...ref.intervention, ...ref.target, ...ref.domain])];
+  const chips = tags
+    .map((tag) => {
+      const stat = tagStats.get(tag);
+      const isUnique = stat.count === 1;
+      const isFirstSeen = !isUnique && ref.year === stat.minYear;
+      let cls = 'tag tag--topic';
+      let flag = '';
+      if (isUnique) {
+        cls += ' tag--unique';
+        flag = '<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span>';
+      } else if (isFirstSeen) {
+        cls += ' tag--new';
+        flag = '<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span>';
+      }
+      return `<span class="${cls}">${esc(tag.replace(/-/g, ' '))}${flag}</span>`;
+    })
+    .join('');
+  return `<div class="reference-entry__meta reference-entry__meta--topics">${chips}</div>`;
+}
+
+function linksRow(ref) {
+  const seen = new Set();
+  const items = [];
+  for (const [key, lbl] of LINK_LABELS) {
+    const url = ref.links[key];
+    if (url && !seen.has(url)) {
+      seen.add(url);
+      items.push(`<a href="${esc(url)}" target="_blank" rel="noopener noreferrer">${lbl}</a>`);
+    }
+  }
+  if (items.length === 0) return '';
+  return `<p class="reference-entry__links" aria-label="Links for ${esc(ref.title)}">${items.join(' &middot; ')}</p>`;
+}
+
+function entryHtml(ref, tagStats) {
+  return `
+          <article class="content-section reference-entry" id="${esc(ref.slug)}">
+            ${tagRow(ref)}
+            <h3>${esc(ref.title)}</h3>
+            <p class="reference-entry__authors">${esc(ref.author.join(', '))}</p>
+            <p>${esc(ref.description)}</p>
+            ${topicTags(ref, tagStats)}
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>${esc(ref.demonstrated)}</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>${esc(ref.ghostati_relevance)}</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>${esc(ref.ghostati_testability.reason)}</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>${esc(ref.limitations)}</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">${esc(ref.citation)} &middot; ${esc(ref.reproducibility.replace(/-/g, ' '))}</p>
+            ${linksRow(ref)}
+          </article>`;
+}
+
+function groupKeyFor(ref) {
+  return GROUP_DEFS.some((d) => d.key === ref.type) ? ref.type : OTHER_GROUP.key;
+}
+
+function buildReferencesBody(sortedRefs, tagStats) {
+  const buckets = new Map();
+  for (const ref of sortedRefs) {
+    const key = groupKeyFor(ref);
+    if (!buckets.has(key)) buckets.set(key, []);
+    buckets.get(key).push(ref);
+  }
+
+  const chunks = [];
+  const activeGroups = [];
+  for (const def of [...GROUP_DEFS, OTHER_GROUP]) {
+    const items = buckets.get(def.key) || [];
+    if (items.length === 0) continue; // don't render (or link to) empty sections
+    activeGroups.push({ ...def, count: items.length });
+    chunks.push(
+      `<h3 class="reference-group" id="${def.anchor}">${esc(def.label)} <span class="reference-group__count">(${items.length})</span></h3>`
+    );
+    chunks.push('<div class="reference-group__list">');
+    for (const ref of items) chunks.push(entryHtml(ref, tagStats));
+    chunks.push('</div>');
+  }
+
+  return { body: chunks.join('\n'), groups: activeGroups };
+}
+
+function main() {
+  // Fail loudly rather than generate a page from a broken/incomplete JSON.
+  try {
+    execFileSync('node', [VALIDATOR_PATH], { stdio: 'inherit', cwd: ROOT });
+  } catch (err) {
+    console.error('\nvalidate-references.js failed — fix REFERENCES.json before building the page.');
+    process.exit(1);
+  }
+
+  const data = JSON.parse(fs.readFileSync(REFERENCES_PATH, 'utf8'));
+  const refs = sortReferences(data.references);
+  const tagStats = buildTagStats(refs);
+
+  const years = refs.map((r) => r.year);
+  const yearMin = Math.min(...years);
+  const yearMax = Math.max(...years);
+  const nResearch = refs.filter((r) => r.type === 'research').length;
+  const nArtistic = refs.filter((r) => r.type === 'artistic').length;
+  const nActivism = refs.filter((r) => r.type === 'activism').length;
+  const nOther = refs.length - nResearch - nArtistic - nActivism;
+
+  const { body, groups } = buildReferencesBody(refs, tagStats);
+  const groupNav = groups.map((g) => `<a href="#${g.anchor}">${esc(g.label)}</a>`).join('\n            ');
+
+  const statsParts = [
+    `${refs.length} references`,
+    `${nResearch} peer-reviewed`,
+    `${nArtistic} art`,
+    `${nActivism} advocacy`,
+  ];
+  if (nOther > 0) statsParts.push(`${nOther} other`);
+  statsParts.push(`${yearMin}&ndash;${yearMax}`);
+  const statsLine = statsParts.join(' &middot; ');
+
+  let html = fs.readFileSync(TEMPLATE_PATH, 'utf8');
+  html = html
+    .replace(/{{REFERENCE_COUNT}}/g, String(refs.length))
+    .replace(/{{YEAR_MIN}}/g, String(yearMin))
+    .replace(/{{YEAR_MAX}}/g, String(yearMax))
+    .replace('{{STATS_LINE}}', statsLine)
+    .replace('{{REFERENCES_BODY}}', body)
+    .replace('{{GROUP_NAV}}', groupNav);
+
+  fs.writeFileSync(OUTPUT_PATH, html);
+  console.log(
+    `Wrote ${path.relative(ROOT, OUTPUT_PATH)}: ${refs.length} entries across ${groups.length} group(s) (${groups
+      .map((g) => `${g.label}: ${g.count}`)
+      .join(', ')}).`
+  );
+}
+
+main();
diff --git a/references/index.html b/references/index.html
new file mode 100644
index 0000000..63ad2df
--- /dev/null
+++ b/references/index.html
@@ -0,0 +1,726 @@
+<!doctype html>
+<html lang="en">
+
+<head>
+  <meta charset="utf-8" />
+  <meta name="viewport" content="width=device-width, initial-scale=1" />
+  <title>References — Ghostmaxxing</title>
+  <meta name="description"
+    content="A curated technical and cultural genealogy of face-recognition camouflage, adversarial makeup, CV Dazzle, face obfuscation and physical-world attacks against computer vision systems." />
+  <link rel="stylesheet" href="/styles/pages.css" />
+  <link rel="stylesheet" href="/styles/content-pages.css" />
+  <link rel="stylesheet" href="/styles/references-list.css" />
+</head>
+
+<body>
+  <!--
+    Relevance-star icon, defined once and reused via <use href="#icon-star">.
+    This is a placeholder star (auto-generated) — replace the single <path>
+    below whenever you have a real icon and every star on the page updates.
+    Keep the id "icon-star" and the 0 0 24 24 viewBox on the <symbol>.
+  -->
+  <svg aria-hidden="true" focusable="false" style="position:absolute;width:0;height:0;overflow:hidden">
+    <symbol id="icon-star" viewBox="0 0 24 24">
+      <path d="M12 2.5 L14.23 8.93 L21.04 9.06 L15.61 13.17 L17.58 19.69 L12 15.8 L6.42 19.69 L8.39 13.17 L2.96 9.06 L9.77 8.93 Z" />
+    </symbol>
+  </svg>
+
+  <main class="editorial-homepage content-page editorial-homepage--svg-footer-band">
+    <div class="editorial-homepage__shell">
+      <header class="editorial-homepage__header">
+        <a class="wordmark" href="index.html" aria-label="Ghostmaxxing homepage">
+          <span class="wordmark__title">Ghostmaxxing</span>
+          <span class="wordmark__subtitle">
+            Privacy activism meets public-research platform
+          </span>
+        </a>
+
+        <nav class="editorial-homepage__nav" aria-label="Project navigation">
+          <a class="navlink" href="https://github.com/vecna/ghostati" target="_blank" rel="noopener noreferrer">Code</a>
+          <a class="navlink" href="about.html">Vision &amp; About</a>
+          <a class="navlink" href="/references" aria-current="page">References</a>
+          <a class="cta-inline" href="report.html">Help by<i>leaking</i>to us<span aria-hidden="true">↗</span></a>
+        </nav>
+      </header>
+
+      <!--
+        Edit this intro freely — it is static and will not be touched by
+        scripts-dev/build-references-page.js. Only the {{...}} placeholders
+        below are replaced by the build script.
+      -->
+      <section class="content-page__header-area" aria-labelledby="page-title">
+        <span class="content-page__kicker">References</span>
+        <h1 id="page-title">A working genealogy of face-recognition camouflage.</h1>
+        <p class="content-page__lead">
+          20 references, 2010&ndash;2025, newest first.
+        </p>
+        <p class="content-page__deck">
+          The Ghostmaxxing references archive maps artistic, academic, and activist work around face obfuscation,
+          adversarial makeup, physical-world attacks, and anti-surveillance aesthetics. Each entry documents what was
+          demonstrated, what its limits are, and whether Ghostmaxxing can directly or partially retest the idea.
+        </p>
+      </section>
+
+      <div class="content-page__layout">
+        <article class="content-page__main">
+          <section class="content-section" id="references-list">
+            <p class="content-page__lead reference-stats">20 references &middot; 15 peer-reviewed &middot; 3 art &middot; 2 advocacy &middot; 2010&ndash;2025</p>
+
+            <h3 class="reference-group" id="group-research">Peer-reviewed papers <span class="reference-group__count">(15)</span></h3>
+<div class="reference-group__list">
+
+          <article class="content-section reference-entry" id="human-imperceptible-nir-face-attack">
+            <div class="reference-entry__meta"><span class="tag">2025</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 1/5 (source score 22/100)"><span class="sr-only">Relevance: 1 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--indirect">Indirect testability</span></div>
+            <h3>Human-Imperceptible Physical Adversarial Attack for NIR Face Recognition Models</h3>
+            <p class="reference-entry__authors">Songyan Xie, Jinghang Wen, Encheng Su, Qiucheng Yu</p>
+            <p>A 2025 physical attack on near-infrared face recognition using infrared-absorbing ink and optimized patches. It is peripheral to Ghostati’s visible-makeup workflow, but important for understanding how anti-biometric interventions move toward sensors and spectra that humans cannot see.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic">infrared</span><span class="tag tag--topic">patch</span><span class="tag tag--topic tag--unique">sensor aware material<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">nir face recognition</span><span class="tag tag--topic">face verification</span><span class="tag tag--topic">physical world</span><span class="tag tag--topic">near infrared</span><span class="tag tag--topic">wearable</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The paper demonstrates a black-box physical adversarial patch for NIR face recognition, using infrared-absorbing ink, skin-reflection modeling, and optimized patch placement to improve attack success in digital and physical domains.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It extends the anti-recognition design space beyond visible cosmetics into NIR-specific materials. Ghostati can use it as a conceptual reference for sensor-aware camouflage, but cannot directly reproduce NIR behavior with a normal browser webcam pipeline.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can document the concept of localized face-region perturbations, but it cannot verify NIR absorption, NIR imaging, or infrared-specific recognition models without dedicated hardware and models.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The reported results are tied to NIR cameras, tested models, patch materials, subject set, and controlled acquisition conditions. Effectiveness against visible-light systems, production deployments, or newer NIR pipelines is not established by this reference.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Xie et al., arXiv 2025 &middot; artifact available</p>
+            <p class="reference-entry__links" aria-label="Links for Human-Imperceptible Physical Adversarial Attack for NIR Face Recognition Models"><a href="https://arxiv.org/abs/2504.15823" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://doi.org/10.48550/arXiv.2504.15823" target="_blank" rel="noopener noreferrer">DOI</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="accessorize-in-the-dark">
+            <div class="reference-entry__meta"><span class="tag">2024</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 1/5 (source score 22/100)"><span class="sr-only">Relevance: 1 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--partial">Partial testability</span></div>
+            <h3>Accessorize in the Dark: A Security Analysis of Near-Infrared Face Recognition</h3>
+            <p class="reference-entry__authors">Amit Cohen, Mahmood Sharif</p>
+            <p>A security analysis of near-infrared face recognition that adapts adversarial accessories to NIR-based systems used in access-control and high-stakes settings.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic">glasses</span><span class="tag tag--topic">near infrared</span><span class="tag tag--topic">wearable accessory</span><span class="tag tag--topic tag--new">nir face recognition<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic">face verification</span><span class="tag tag--topic">physical world</span><span class="tag tag--topic">wearable</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The work tests NIR-based face recognition against physically realizable adversarial accessories, reporting high physical attack success across defended and undefended models.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It is relevant because it shows that face-recognition fragility is not limited to visible-light cameras or RGB makeup. It helps Ghostati explain why different sensors require different camouflage strategies.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can approximate the visible placement of a wearable accessory around the face, but cannot test NIR capture, NIR-specific perturbations, or defended NIR models in the browser.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The result depends on NIR capture assumptions, model set, accessory realization, and experimental protocol. A standard RGB webcam test in Ghostati would not confirm the NIR-specific claim.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Cohen and Sharif, ESORICS 2023 / LNCS 2024 &middot; code available</p>
+            <p class="reference-entry__links" aria-label="Links for Accessorize in the Dark: A Security Analysis of Near-Infrared Face Recognition"><a href="https://doi.org/10.1007/978-3-031-51479-1_3" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://link.springer.com/chapter/10.1007/978-3-031-51479-1_3" target="_blank" rel="noopener noreferrer">Paper</a> &middot; <a href="https://github.com/AmitCohen3/Accessorize-in-the-dark" target="_blank" rel="noopener noreferrer">Code</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="physical-world-optical-attacks-3d-face">
+            <div class="reference-entry__meta"><span class="tag">2023</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 1/5 (source score 30/100)"><span class="sr-only">Relevance: 1 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--indirect">Indirect testability</span></div>
+            <h3>Physical-World Optical Adversarial Attacks on 3D Face Recognition</h3>
+            <p class="reference-entry__authors">Yanjie Li, Yiquan Li, Xuelong Dai, Songtao Guo, Bin Xiao</p>
+            <p>A CVPR 2023 paper on structured-light optical attacks against 3D face recognition systems, moving anti-recognition intervention from visible surface patterns to depth reconstruction and projected perturbations.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic tag--unique">structured light<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">optical</span><span class="tag tag--topic">projection</span><span class="tag tag--topic tag--unique">3d face recognition<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">depth face recognition<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">physical world</span><span class="tag tag--topic tag--unique">3d<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The paper demonstrates a structured-light attack that can affect point-cloud-based and depth-image-based 3D face recognition with fewer perturbations and resilience to random head movements.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It helps Ghostati explain that face recognition is not only 2D RGB matching: depth, structured light, and skin reflectance introduce a different attack surface.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati’s browser pipeline can discuss optical perturbations conceptually, but cannot reproduce structured-light 3D scanning, depth-image recognition, or point-cloud attacks.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The setup requires structured-light 3D recognition assumptions and specialized acquisition. It is not evidence that ordinary visible AR makeup affects 3D systems in the same way.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Li et al., CVPR 2023 &middot; paper available</p>
+            <p class="reference-entry__links" aria-label="Links for Physical-World Optical Adversarial Attacks on 3D Face Recognition"><a href="https://openaccess.thecvf.com/content/CVPR2023/html/Li_Physical-World_Optical_Adversarial_Attacks_on_3D_Face_Recognition_CVPR_2023_paper.html" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://openaccess.thecvf.com/content/CVPR2023/papers/Li_Physical-World_Optical_Adversarial_Attacks_on_3D_Face_Recognition_CVPR_2023_paper.pdf" target="_blank" rel="noopener noreferrer">Paper</a> &middot; <a href="https://github.com/PolyLiYJ/SLAttack" target="_blank" rel="noopener noreferrer">Code</a> &middot; <a href="https://arxiv.org/abs/2205.13412" target="_blank" rel="noopener noreferrer">arXiv</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="shadows-can-be-dangerous">
+            <div class="reference-entry__meta"><span class="tag">2022</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 1/5 (source score 35/100)"><span class="sr-only">Relevance: 1 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--contextual">Contextual reference</span></div>
+            <h3>Shadows can be Dangerous: Stealthy and Effective Physical-world Adversarial Attack by Natural Phenomenon</h3>
+            <p class="reference-entry__authors">Yiqi Zhong, Xianming Liu, Deming Zhai, Junjun Jiang, Xiangyang Ji</p>
+            <p>A physical-world adversarial attack that uses natural-looking shadows rather than stickers or conspicuous projected patterns, tested on traffic-sign recognition rather than face recognition.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic">shadow</span><span class="tag tag--topic">optical</span><span class="tag tag--topic tag--unique">environmental perturbation<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">traffic sign recognition<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">image classification</span><span class="tag tag--topic">physical world</span><span class="tag tag--topic">visible light</span><span class="tag tag--topic tag--unique">environmental<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The paper demonstrates that optimized shadows can create stealthy black-box physical adversarial examples, with strong simulated and real-world attack results on traffic-sign datasets.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It is useful to Ghostati as a design reference for natural phenomena, lighting, and environmental perturbations, but it is not face-specific.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can use this as a conceptual reference for light and shadow as adversarial surfaces, but cannot validate the traffic-sign attack or transfer it to face recognition without a new experiment.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The demonstrated target is traffic-sign recognition, not face detection or recognition. Any relevance to Ghostati requires cautious analogy, not direct transfer of the result.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Zhong et al., CVPR 2022 &middot; code available</p>
+            <p class="reference-entry__links" aria-label="Links for Shadows can be Dangerous: Stealthy and Effective Physical-world Adversarial Attack by Natural Phenomenon"><a href="https://arxiv.org/abs/2203.03818" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://github.com/hncszyq/ShadowAttack" target="_blank" rel="noopener noreferrer">Code</a> &middot; <a href="https://doi.org/10.48550/arXiv.2203.03818" target="_blank" rel="noopener noreferrer">DOI</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="adv-makeup">
+            <div class="reference-entry__meta"><span class="tag">2021</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 4/5 (source score 95/100)"><span class="sr-only">Relevance: 4 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--partial">Partial testability</span></div>
+            <h3>Adv-Makeup: A New Imperceptible and Transferable Attack on Face Recognition</h3>
+            <p class="reference-entry__authors">Bangjie Yin, Wenxuan Wang, Taiping Yao, Junfeng Guo, Zelun Kong, Shouhong Ding, Jilin Li, Cong Liu</p>
+            <p>A key adversarial-makeup paper that synthesizes transferable, imperceptible eye-shadow perturbations over the orbital region to attack face recognition.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic">makeup</span><span class="tag tag--topic tag--unique">orbital region<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">adversarial optimization</span><span class="tag tag--topic">face recognition</span><span class="tag tag--topic">face verification</span><span class="tag tag--topic">physical world</span><span class="tag tag--topic">digital</span><span class="tag tag--topic">visible light</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The paper demonstrates a task-driven makeup generation method and fine-grained meta-learning strategy intended to improve black-box transferability, including digital and physical scenarios.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>This is one of the closest research references for Ghostati because it treats makeup as an adversarial surface on the machine-readable face.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can approximate orbital-region overlays and compare local recognition behavior, but it does not implement the paper’s adversarial optimization, meta-learning, or commercial-system evaluation protocol.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The reported performance depends on the models, datasets, optimization method, physical-printing or application process, and commercial APIs tested in the paper. A manually designed Ghostyle is not equivalent to Adv-Makeup optimization.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Yin et al., IJCAI 2021 &middot; paper available</p>
+            <p class="reference-entry__links" aria-label="Links for Adv-Makeup: A New Imperceptible and Transferable Attack on Face Recognition"><a href="https://arxiv.org/abs/2105.03162" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://doi.org/10.48550/arXiv.2105.03162" target="_blank" rel="noopener noreferrer">DOI</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="adversarial-attacks-against-face-recognition">
+            <div class="reference-entry__meta"><span class="tag">2020</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 3/5 (source score 82/100)"><span class="sr-only">Relevance: 3 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--contextual">Contextual reference</span></div>
+            <h3>Adversarial Attacks against Face Recognition: A Comprehensive Study</h3>
+            <p class="reference-entry__authors">Fatemeh Vakhshiteh, Ahmad Nickabadi, Raghavendra Ramachandra</p>
+            <p>A broad survey of adversarial attacks and defenses against face recognition systems, useful for situating Ghostati within a larger technical taxonomy.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic tag--unique">survey<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">taxonomy<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">face recognition</span><span class="tag tag--topic">face verification</span><span class="tag tag--topic">digital</span><span class="tag tag--topic">physical world</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The article surveys attack and defense methods, proposes taxonomies, compares criteria, and identifies challenges and future research directions for adversarial face recognition.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It gives Ghostati a technical map for classifying references by threat model, physical/digital domain, target stage, and defense assumptions.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can use the survey to classify experiments and limitations, but a survey is not itself a system or attack to retest.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>As a survey, it summarizes prior work rather than providing one reproducible experimental setup. Some included systems and defenses may have changed since publication.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Vakhshiteh, Nickabadi, and Ramachandra, 2020/2021 &middot; survey</p>
+            <p class="reference-entry__links" aria-label="Links for Adversarial Attacks against Face Recognition: A Comprehensive Study"><a href="https://arxiv.org/abs/2007.11709" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://doi.org/10.48550/arXiv.2007.11709" target="_blank" rel="noopener noreferrer">DOI</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="breaking-certified-defenses-shadow-attack">
+            <div class="reference-entry__meta"><span class="tag">2020</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 1/5 (source score 35/100)"><span class="sr-only">Relevance: 1 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--contextual">Contextual reference</span></div>
+            <h3>Breaking certified defenses: Semantic adversarial examples with spoofed robustness certificates</h3>
+            <p class="reference-entry__authors">Amin Ghiasi, Ali Shafahi, Tom Goldstein</p>
+            <p>A semantic adversarial-example paper that uses large but perceptually plausible perturbations to attack robust classifiers and spoof robustness certificates.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic tag--new">shadow<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--unique">semantic perturbation<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">image classification</span><span class="tag tag--topic tag--unique">certified defense<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">digital</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The paper demonstrates a Shadow Attack that can cause certified classifiers to mislabel images while still producing a spoofed robustness certificate.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It is conceptually relevant because it frames adversarial design as a semantic, human-plausible transformation rather than only pixel noise, but it is not face- or makeup-specific.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can borrow the idea of perceptually plausible visual transformations, but cannot retest certified robustness or the paper’s classifier/certificate setting.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The work is not face-specific and addresses certified image classifiers rather than browser face matching. Relevance to Ghostati is conceptual and should not be presented as evidence for face-camouflage effectiveness.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Ghiasi, Shafahi, and Goldstein, 2020 &middot; paper available</p>
+            <p class="reference-entry__links" aria-label="Links for Breaking certified defenses: Semantic adversarial examples with spoofed robustness certificates"><a href="https://arxiv.org/abs/2003.08937" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://doi.org/10.48550/arXiv.2003.08937" target="_blank" rel="noopener noreferrer">DOI</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="vla-visible-light-face-attack">
+            <div class="reference-entry__meta"><span class="tag">2019</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 3/5 (source score 78/100)"><span class="sr-only">Relevance: 3 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--partial">Partial testability</span></div>
+            <h3>VLA: A Practical Visible Light-based Attack on Face Recognition Systems in Physical World</h3>
+            <p class="reference-entry__authors">Meng Shen, Zelin Liao, Liehuang Zhu, Ke Xu, Xiaojiang Du</p>
+            <p>A visible-light physical attack that projects adversarial perturbations onto a face to conduct targeted or untargeted attacks against black-box face recognition systems.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic tag--unique">light<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--new">projection<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--new">optical<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic">face recognition</span><span class="tag tag--topic">face verification</span><span class="tag tag--topic">physical world</span><span class="tag tag--topic">visible light</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The paper proposes VLA, a visible-light projection method for crafting physical adversarial perturbations on the face against black-box face recognition systems.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It connects Ghostati’s visual overlay approach to optical projection attacks: both change the appearance captured by the camera, but VLA is optimized and physically projected rather than drawn as a browser layer.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can simulate a visible overlay on the face and measure local matching changes, but it cannot validate projector optics, camera-image formation, or the paper’s black-box attack setup.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The result depends on projector/camera geometry, brightness, environmental lighting, attack optimization, and the target recognition systems. Browser AR overlays are not equivalent to physically projected light.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Shen et al., UbiComp/IMWUT 2019 &middot; paper available</p>
+            <p class="reference-entry__links" aria-label="Links for VLA: A Practical Visible Light-based Attack on Face Recognition Systems in Physical World"><a href="https://doi.org/10.1145/3351261" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://www.thucsnet.com/wp-content/papers/meng_ubicomp19.pdf" target="_blank" rel="noopener noreferrer">Paper</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="art-toolbox">
+            <div class="reference-entry__meta"><span class="tag">2018</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 2/5 (source score 55/100)"><span class="sr-only">Relevance: 2 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--indirect">Indirect testability</span></div>
+            <h3>Adversarial Robustness Toolbox v1.0.0</h3>
+            <p class="reference-entry__authors">Maria-Irina Nicolae, Mathieu Sinn, Minh Ngoc Tran, Beat Buesser, Ambrish Rawat, Martin Wistuba, Valentina Zantedeschi, Nathalie Baracaldo, Bryant Chen, Heiko Ludwig, Ian M. Molloy, Ben Edwards</p>
+            <p>A general-purpose Python library for testing machine-learning models with adversarial attacks and defenses, useful as technical infrastructure rather than as a face-camouflage reference.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic tag--unique">toolkit<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">adversarial ml<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">machine learning models<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">digital</span><span class="tag tag--topic tag--unique">software library<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>ART provides implementations, examples, and documentation for attacks and defenses across many model families and machine-learning frameworks.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It helps situate Ghostati inside adversarial ML practice, but it is not a visual face-camouflage project by itself.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati could conceptually integrate adversarial-evaluation workflows, but ART does not directly map to the current browser-only Ghostyle test loop.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>ART is broad infrastructure, not a face-specific or physical-world camouflage result. Applying it to Ghostati would require additional model wrappers and experiment design.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Nicolae et al., 2018/2019 &middot; code available</p>
+            <p class="reference-entry__links" aria-label="Links for Adversarial Robustness Toolbox v1.0.0"><a href="https://arxiv.org/abs/1807.01069" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://github.com/Trusted-AI/adversarial-robustness-toolbox" target="_blank" rel="noopener noreferrer">Code</a> &middot; <a href="https://doi.org/10.48550/arXiv.1807.01069" target="_blank" rel="noopener noreferrer">DOI</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="dpatch">
+            <div class="reference-entry__meta"><span class="tag">2018</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 2/5 (source score 40/100)"><span class="sr-only">Relevance: 2 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--indirect">Indirect testability</span></div>
+            <h3>DPatch: An Adversarial Patch Attack on Object Detectors</h3>
+            <p class="reference-entry__authors">Xin Liu, Huanrui Yang, Ziwei Liu, Linghao Song, Hai Li, Yiran Chen</p>
+            <p>A black-box adversarial patch attack against object detectors such as Faster R-CNN and YOLO, relevant to the broader genealogy of physical adversarial patches.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic">patch</span><span class="tag tag--topic">adversarial optimization</span><span class="tag tag--topic tag--new">object detection<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic">digital</span><span class="tag tag--topic">physical world</span><span class="tag tag--topic">printed</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>DPatch demonstrates that a small, location-independent adversarial patch can attack both bounding-box regression and object classification in object detectors, with transferability across detectors.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It is not face-specific, but it explains why localized graphic surfaces can disrupt detection pipelines, a core idea behind many physical-world adversarial designs.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can draw face-localized patches, but DPatch targets object detectors and requires detector-specific optimization that Ghostati does not implement.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The attack target is object detection, not face recognition. Transfer to face analysis is conceptual unless a new face-specific optimization and evaluation are performed.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Liu et al., SafeAI 2019 / arXiv 2018 &middot; paper available</p>
+            <p class="reference-entry__links" aria-label="Links for DPatch: An Adversarial Patch Attack on Object Detectors"><a href="https://arxiv.org/abs/1806.02299" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://doi.org/10.48550/arXiv.1806.02299" target="_blank" rel="noopener noreferrer">DOI</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="shapeshifter">
+            <div class="reference-entry__meta"><span class="tag">2018</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 1/5 (source score 38/100)"><span class="sr-only">Relevance: 1 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--contextual">Contextual reference</span></div>
+            <h3>ShapeShifter: Robust Physical Adversarial Attack on Faster R-CNN Object Detector</h3>
+            <p class="reference-entry__authors">Shang-Tse Chen, Cory Cornelius, Jason Martin, Duen Horng Chau</p>
+            <p>A physical adversarial attack on Faster R-CNN object detection that adapts robust transformation methods to real-world detector attacks.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic">patch</span><span class="tag tag--topic">printed</span><span class="tag tag--topic">adversarial optimization</span><span class="tag tag--topic tag--new">object detection<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic">physical world</span><span class="tag tag--topic">visible light</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>ShapeShifter shows that physical perturbations can be optimized to survive real-world transformations such as distance, angle, lighting, and camera limitations in an object-detection setting.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It is a foundational physical-world adversarial reference: not face-centered, but important for understanding robustness under real-world transformations.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can borrow its concern with transformations, but cannot directly retest Faster R-CNN stop-sign attacks or autonomous-driving scenarios.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The target domain is object detection, not biometric recognition. Its lessons transfer as methodology, not as a direct Ghostati result.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Chen et al., ECML PKDD 2018 &middot; paper available</p>
+            <p class="reference-entry__links" aria-label="Links for ShapeShifter: Robust Physical Adversarial Attack on Faster R-CNN Object Detector"><a href="https://arxiv.org/abs/1804.05810" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://doi.org/10.1007/978-3-030-10925-7_4" target="_blank" rel="noopener noreferrer">DOI</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="agns-face-recognition">
+            <div class="reference-entry__meta"><span class="tag">2017</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 4/5 (source score 90/100)"><span class="sr-only">Relevance: 4 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--partial">Partial testability</span></div>
+            <h3>A General Framework for Adversarial Examples with Objectives</h3>
+            <p class="reference-entry__authors">Mahmood Sharif, Sruti Bhagavatula, Lujo Bauer, Michael K. Reiter</p>
+            <p>The later framework paper behind adversarial generative nets, including physical eyeglass frames designed to fool face recognition with better robustness, inconspicuousness, and scalability than earlier approaches.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic">glasses</span><span class="tag tag--topic">wearable accessory</span><span class="tag tag--topic">adversarial optimization</span><span class="tag tag--topic">face recognition</span><span class="tag tag--topic">face verification</span><span class="tag tag--topic">physical world</span><span class="tag tag--topic">visible light</span><span class="tag tag--topic">wearable</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The paper demonstrates adversarial generative nets that can satisfy application-specific objectives, including physical eyeglass-frame examples for face recognition.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It is highly relevant because it treats facial accessories as optimized physical adversarial media, adjacent to Ghostati’s Ghostyle concept even though Ghostati uses browser-rendered visual styles rather than generated eyeglass frames.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can retest the broad idea of face-region or accessory-like overlays, but not the generator training, eyeglass printing pipeline, or exact attack objective.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The result depends on generated eyeglass-frame patterns, model assumptions, and physical realization. Browser overlays can approximate placement but not validate the original optimization or robustness claims.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Sharif et al., arXiv 2017 / ACM TOPS 2019 &middot; paper available</p>
+            <p class="reference-entry__links" aria-label="Links for A General Framework for Adversarial Examples with Objectives"><a href="https://arxiv.org/abs/1801.00349" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://doi.org/10.1145/3317611" target="_blank" rel="noopener noreferrer">DOI</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="adversarial-patch">
+            <div class="reference-entry__meta"><span class="tag">2017</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 2/5 (source score 50/100)"><span class="sr-only">Relevance: 2 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--indirect">Indirect testability</span></div>
+            <h3>Adversarial Patch</h3>
+            <p class="reference-entry__authors">Tom B. Brown, Dandelion Mané, Aurko Roy, Martín Abadi, Justin Gilmer</p>
+            <p>A seminal paper on universal, robust, targeted adversarial image patches that can be printed, added to scenes, photographed, and presented to classifiers.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic tag--new">patch<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--new">printed<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic">adversarial optimization</span><span class="tag tag--topic">image classification</span><span class="tag tag--topic">physical world</span><span class="tag tag--topic">visible light</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The work demonstrates printable universal patches that can cause classifiers to ignore other scene content and output a chosen target class across transformations.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It provides the core technical idea behind many visible adversarial surfaces: a localized graphic object can dominate a model’s interpretation of an image.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can draw patch-like overlays, but the original paper targets image classifiers and requires optimization not present in Ghostati.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The paper is not face-specific. Its relevance to Ghostati is methodological unless a face-recognition-specific patch is optimized and tested.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Brown et al., 2017/2018 &middot; code available</p>
+            <p class="reference-entry__links" aria-label="Links for Adversarial Patch"><a href="https://arxiv.org/abs/1712.09665" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://github.com/tensorflow/cleverhans/tree/master/examples/adversarial_patch" target="_blank" rel="noopener noreferrer">Code</a> &middot; <a href="https://doi.org/10.48550/arXiv.1712.09665" target="_blank" rel="noopener noreferrer">DOI</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="accessorize-to-a-crime">
+            <div class="reference-entry__meta"><span class="tag">2016</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 4/5 (source score 92/100)"><span class="sr-only">Relevance: 4 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--partial">Partial testability</span></div>
+            <h3>Accessorize to a Crime: Real and Stealthy Attacks on State-of-the-Art Face Recognition</h3>
+            <p class="reference-entry__authors">Mahmood Sharif, Sruti Bhagavatula, Lujo Bauer, Michael K. Reiter</p>
+            <p>A landmark physical attack on face recognition using printed eyeglass frames to support evasion or impersonation attacks in a comparatively inconspicuous form.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic tag--new">glasses<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--new">wearable accessory<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--new">adversarial optimization<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--new">face recognition<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--new">face verification<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic">physical world</span><span class="tag tag--topic">visible light</span><span class="tag tag--topic tag--new">wearable<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The paper defines physically realizable and inconspicuous attacks on facial biometric systems and realizes them by printing adversarial eyeglass frames.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It is one of the strongest technical ancestors for Ghostati’s appearance-based tests: the face is not hidden, but visually modified to change machine matching.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can approximate glasses-like overlays and compare local matching behavior, but cannot reproduce printed-material effects, physical-camera capture, or the exact optimization pipeline.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The results are tied to the target systems, printed patterns, imaging conditions, and experimental subjects. Newer recognition models or production systems require fresh testing.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Sharif et al., ACM CCS 2016 &middot; code available</p>
+            <p class="reference-entry__links" aria-label="Links for Accessorize to a Crime: Real and Stealthy Attacks on State-of-the-Art Face Recognition"><a href="https://doi.org/10.1145/2976749.2978392" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://users.ece.cmu.edu/~lbauer/papers/2016/ccs2016-face-recognition.pdf" target="_blank" rel="noopener noreferrer">Paper</a> &middot; <a href="https://mahmoods01.github.io/publication/ccs16-adv-ml/" target="_blank" rel="noopener noreferrer">Project</a> &middot; <a href="https://github.com/mahmoods01/accessorize-to-a-crime" target="_blank" rel="noopener noreferrer">Code</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="adversarial-manipulation-deep-representations">
+            <div class="reference-entry__meta"><span class="tag">2015</span><span class="tag">Research</span><span class="reference-relevance" title="Relevance 2/5 (source score 45/100)"><span class="sr-only">Relevance: 2 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--contextual">Contextual reference</span></div>
+            <h3>Adversarial Manipulation of Deep Representations</h3>
+            <p class="reference-entry__authors">Sara Sabour, Yanshuai Cao, Fartash Faghri, David J. Fleet</p>
+            <p>A theoretical reference showing that an image’s internal deep-network representation can be manipulated to resemble another image while remaining visually similar to the original.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic tag--unique">digital perturbation<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">representation attack<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">deep representation<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--new">image classification<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--new">digital<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The paper demonstrates feature adversaries: small perturbations that make internal DNN representations mimic different natural images, raising questions about what representations encode.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It gives Ghostati a conceptual basis for thinking beyond labels: recognition systems operate through internal representations, not human-readable identity alone.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can discuss representation shifts, but it does not expose or optimize internal DNN layers in the current browser pipeline.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>This is not a physical-world or face-specific intervention. Its connection to Ghostati is conceptual unless the project adds model-internal representation testing.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Sabour et al., ICLR 2016 &middot; paper available</p>
+            <p class="reference-entry__links" aria-label="Links for Adversarial Manipulation of Deep Representations"><a href="https://arxiv.org/abs/1511.05122" target="_blank" rel="noopener noreferrer">Canonical</a> &middot; <a href="https://doi.org/10.48550/arXiv.1511.05122" target="_blank" rel="noopener noreferrer">DOI</a></p>
+          </article>
+</div>
+<h3 class="reference-group" id="group-artistic">Art initiatives <span class="reference-group__count">(3)</span></h3>
+<div class="reference-group__list">
+
+          <article class="content-section reference-entry" id="camera-shy-hoodie">
+            <div class="reference-entry__meta"><span class="tag">2023</span><span class="tag">Artistic</span><span class="reference-relevance" title="Relevance 1/5 (source score 20/100)"><span class="sr-only">Relevance: 1 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--no">Not testable in-browser</span></div>
+            <h3>The Camera-Shy Hoodie</h3>
+            <p class="reference-entry__authors">Mac Pierce</p>
+            <p>A DIY wearable that uses high-power infrared LEDs to degrade or overexpose night-vision security camera footage around the wearer’s head and upper body.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic tag--new">infrared<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--unique">garment<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">sensor jamming<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">security camera<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">sensor disturbance<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">physical world</span><span class="tag tag--topic tag--new">near infrared<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic">wearable</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The project demonstrates a buildable garment that interferes with IR-sensitive security cameras by using the same infrared wavelength commonly used for night-vision illumination.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It is a useful boundary reference: not face makeup and not face recognition in the strict sense, but a clear example of physical sensor disturbance as anti-surveillance design.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati cannot test IR LED glare, camera auto-exposure interference, or night-vision capture in a standard browser AR pipeline.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>Its effect depends on camera sensitivity, exposure behavior, LED placement, distance, and lighting. It does not directly test face-recognition matching or browser-based face analysis.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Mac Pierce, The Camera-Shy Hoodie, 2023 &middot; build files available</p>
+            <p class="reference-entry__links" aria-label="Links for The Camera-Shy Hoodie"><a href="https://www.macpierce.com/the-camera-shy-hoodie" target="_blank" rel="noopener noreferrer">Canonical</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="hyperface">
+            <div class="reference-entry__meta"><span class="tag">2016</span><span class="tag">Artistic</span><span class="reference-relevance" title="Relevance 4/5 (source score 92/100)"><span class="sr-only">Relevance: 4 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--partial">Partial testability</span></div>
+            <h3>HyperFace</h3>
+            <p class="reference-entry__authors">Adam Harvey</p>
+            <p>Adam Harvey’s false-face textile camouflage, developed with Hyphen-Labs, that attempts to redirect face-detection confidence toward surrounding decoy faces rather than the wearer’s real face.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic tag--unique">textile<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">false face<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">decoy<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">pattern<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">face detection</span><span class="tag tag--topic">physical world</span><span class="tag tag--topic">visible light</span><span class="tag tag--topic tag--new">wearable<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>HyperFace demonstrates a ground-based camouflage strategy: instead of hiding the face directly, it supplies machine-readable false faces around it to distract specific detectors.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It expands Ghostati’s design vocabulary from face overlays to surrounding false-face patterns and attention redirection.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can test surrounding decoy-face graphics against browser detection in a limited way, but the original HyperFace patterns targeted specific Haarcascade detectors and textile contexts.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The documented prototype targeted OpenCV Haarcascade-style detectors and is not designed or tested against modern deep CNN-based face detection systems. Its value as camouflage is explicitly temporary and system-specific.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Adam Harvey, HyperFace, 2016/2017 &middot; documented project</p>
+            <p class="reference-entry__links" aria-label="Links for HyperFace"><a href="https://adam.harvey.studio/hyperface/" target="_blank" rel="noopener noreferrer">Canonical</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="cv-dazzle">
+            <div class="reference-entry__meta"><span class="tag">2010</span><span class="tag">Artistic</span><span class="reference-relevance" title="Relevance 5/5 (source score 100/100)"><span class="sr-only">Relevance: 5 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--direct">Direct testability</span></div>
+            <h3>CV Dazzle</h3>
+            <p class="reference-entry__authors">Adam Harvey</p>
+            <p>Adam Harvey’s foundational computer-vision camouflage project, created in 2010, using makeup, hairstyling, and asymmetry to disrupt face detection while remaining visible to humans.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic tag--new">makeup<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--unique">hair<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">asymmetry<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--new">face obfuscation<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--new">face detection<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--new">physical world<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--new">visible light<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>CV Dazzle demonstrated that hair and makeup alone could lower detection probability below threshold for several OpenCV Haarcascade face-detection profiles in the original proof-of-concept setting.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It is the central historical anchor for Ghostati: a technique-first, system-specific approach to anti-biometric appearance design.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can retest the general strategy of visible face-region interventions against its own browser detection and matching pipeline, while clearly separating old Haarcascade claims from modern models.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The original looks were designed for Viola-Jones/OpenCV Haarcascade detectors and should not be treated as active protection against modern CNN-based detection or recognition systems. Lighting, pose, algorithm, and wearer-specific design all matter.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Adam Harvey, CV Dazzle, 2010 &middot; documented project</p>
+            <p class="reference-entry__links" aria-label="Links for CV Dazzle"><a href="https://adam.harvey.studio/cvdazzle/" target="_blank" rel="noopener noreferrer">Canonical</a></p>
+          </article>
+</div>
+<h3 class="reference-group" id="group-activism">Advocacy efforts <span class="reference-group__count">(2)</span></h3>
+<div class="reference-group__list">
+
+          <article class="content-section reference-entry" id="dazzle">
+            <div class="reference-entry__meta"><span class="tag">2023</span><span class="tag">Activism</span><span class="reference-relevance" title="Relevance 5/5 (source score 100/100)"><span class="sr-only">Relevance: 5 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--direct">Direct testability</span></div>
+            <h3>DAZZLE</h3>
+            <p class="reference-entry__authors">Michelle Tylicki, Lauri Love</p>
+            <p>A collaborative art-and-technology experiment by Michelle Tylicki and Lauri Love that turns anti-surveillance makeup into an interactive teaching and testing experience.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic">makeup</span><span class="tag tag--topic">face obfuscation</span><span class="tag tag--topic tag--unique">interactive tool<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">workshop</span><span class="tag tag--topic">face detection</span><span class="tag tag--topic">face recognition</span><span class="tag tag--topic">physical world</span><span class="tag tag--topic tag--unique">browser<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">visible light</span><span class="tag tag--topic">festival</span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The project demonstrates a salon-like interface for teaching Computer Vision Dazzle and testing camouflage against multiple facial recognition systems in an activist and public-learning context.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>DAZZLE is one of Ghostati’s closest contemporary relatives: it combines makeup, interface design, quantitative feedback, and anti-surveillance pedagogy.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can directly retest the browser-lab logic of applying face-region visual interventions and comparing detection or matching behavior before and after the intervention.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>Public documentation describes the project and its testing ambition, but does not provide a full reproducible benchmark, exact model list, or protocol. Claims should be read as project documentation unless the implementation and results are independently released.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">Tylicki and Love, DAZZLE, 2023 &middot; documented project</p>
+            <p class="reference-entry__links" aria-label="Links for DAZZLE"><a href="https://www.michelletylicki.info/dazzle/" target="_blank" rel="noopener noreferrer">Canonical</a></p>
+          </article>
+
+          <article class="content-section reference-entry" id="dazzle-club">
+            <div class="reference-entry__meta"><span class="tag">2019</span><span class="tag">Activism</span><span class="reference-relevance" title="Relevance 4/5 (source score 96/100)"><span class="sr-only">Relevance: 4 out of 5</span><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg><svg class="star-icon star-icon--empty" width="14" height="14" viewBox="0 0 24 24" aria-hidden="true"><use href="#icon-star" xlink:href="#icon-star"></use></svg></span><span class="badge badge--contextual">Contextual reference</span></div>
+            <h3>The Dazzle Club</h3>
+            <p class="reference-entry__authors">Evie Price, Emily Roderick, Georgina Rowlands, Anna Hart</p>
+            <p>A London-based collective practice using CV Dazzle walks, workshops, films, and public-space research to draw attention to facial recognition and surveillance infrastructure.</p>
+            <div class="reference-entry__meta reference-entry__meta--topics"><span class="tag tag--topic">makeup</span><span class="tag tag--topic tag--unique">performance<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--unique">public walk<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--new">workshop<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span><span class="tag tag--topic tag--unique">public space surveillance<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic">face recognition</span><span class="tag tag--topic">physical world</span><span class="tag tag--topic tag--unique">public space<span class="tag__flag" title="Only this entry in the archive uses this tag">only here</span></span><span class="tag tag--topic tag--new">festival<span class="tag__flag" title="Earliest entry in the archive using this tag">first seen</span></span></div>
+            <details class="reference-entry__assessment">
+              <summary>Full assessment</summary>
+              <dl class="reference-entry__facts">
+                <div class="reference-entry__fact">
+                  <dt>Demonstrated</dt>
+                  <dd>The Dazzle Club demonstrates how face paint and CV Dazzle can operate as embodied public research, collective performance, and public education rather than only as a technical evasion method.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Why it matters for Ghostmaxxing</dt>
+                  <dd>It helps Ghostati understand camouflage as a social and workshop practice: a way to make public surveillance visible, discussable, and collectively testable.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Ghostmaxxing testability</dt>
+                  <dd>Ghostati can support similar workshops and local tests, but it cannot reproduce the public-space performance, legal context, or collective embodied research of the project.</dd>
+                </div>
+                <div class="reference-entry__fact">
+                  <dt>Limits</dt>
+                  <dd>The project is primarily an artistic and activist practice, not a controlled technical benchmark. Its value is public research and awareness rather than a reproducible recognition-evasion result.</dd>
+                </div>
+              </dl>
+            </details>
+            <p class="reference-entry__citation">The Dazzle Club, 2019–2021 &middot; artwork documentation</p>
+            <p class="reference-entry__links" aria-label="Links for The Dazzle Club"><a href="https://emilyroderick.com/work/the-dazzle-club/" target="_blank" rel="noopener noreferrer">Canonical</a></p>
+          </article>
+</div>
+          </section>
+
+          <!-- Static section — edit freely -->
+          <section class="content-section" id="method">
+            <h2>How to suggest a new entry.</h2>
+            <p>
+              Every possible referenes should be read and transformed into the format of <code>REFERENCES.json</code>, check out on github (https://github.com/vecna/ghostati/blob/rebranding/ghostmaxxing/references/REFERENCES.json) how detailed is.
+              Once you're sure it might suit us, and potentially we can even program a Ghostyle to retest it, please open an issue or submit a pull request.
+            </p>
+          </section>
+        </article>
+
+        <aside class="content-page__aside" aria-label="Page sections">
+          <h2>Jump to</h2>
+          <nav>
+            <a href="#group-research">Peer-reviewed papers</a>
+            <a href="#group-artistic">Art initiatives</a>
+            <a href="#group-activism">Advocacy efforts</a>
+            <a href="#method">Suggest a new entry</a>
+          </nav>
+        </aside>
+      </div>
+    </div>
+
+    <div class="camera-band" aria-hidden="true"></div>
+  </main>
+</body>
+
+</html>
diff --git a/references/templates/references.template.html b/references/templates/references.template.html
new file mode 100644
index 0000000..43a05c8
--- /dev/null
+++ b/references/templates/references.template.html
@@ -0,0 +1,101 @@
+<!doctype html>
+<html lang="en">
+
+<head>
+  <meta charset="utf-8" />
+  <meta name="viewport" content="width=device-width, initial-scale=1" />
+  <title>References — Ghostmaxxing</title>
+  <meta name="description"
+    content="A curated technical and cultural genealogy of face-recognition camouflage, adversarial makeup, CV Dazzle, face obfuscation and physical-world attacks against computer vision systems." />
+  <link rel="stylesheet" href="/styles/pages.css" />
+  <link rel="stylesheet" href="/styles/content-pages.css" />
+  <link rel="stylesheet" href="/styles/references-list.css" />
+</head>
+
+<body>
+  <!--
+    Relevance-star icon, defined once and reused via <use href="#icon-star">.
+    This is a placeholder star (auto-generated) — replace the single <path>
+    below whenever you have a real icon and every star on the page updates.
+    Keep the id "icon-star" and the 0 0 24 24 viewBox on the <symbol>.
+  -->
+  <svg aria-hidden="true" focusable="false" style="position:absolute;width:0;height:0;overflow:hidden">
+    <symbol id="icon-star" viewBox="0 0 24 24">
+      <path
+        d="M12 2.5 L14.23 8.93 L21.04 9.06 L15.61 13.17 L17.58 19.69 L12 15.8 L6.42 19.69 L8.39 13.17 L2.96 9.06 L9.77 8.93 Z" />
+    </symbol>
+  </svg>
+
+  <main class="editorial-homepage content-page editorial-homepage--svg-footer-band">
+    <div class="editorial-homepage__shell">
+      <header class="editorial-homepage__header">
+        <a class="wordmark" href="index.html" aria-label="Ghostmaxxing homepage">
+          <span class="wordmark__title">Ghostmaxxing</span>
+          <span class="wordmark__subtitle">
+            Privacy activism meets public-research platform
+          </span>
+        </a>
+
+        <nav class="editorial-homepage__nav" aria-label="Project navigation">
+          <a class="navlink" href="https://github.com/vecna/ghostati" target="_blank" rel="noopener noreferrer">Code</a>
+          <a class="navlink" href="about.html">Vision &amp; About</a>
+          <a class="navlink" href="/references" aria-current="page">References</a>
+          <a class="cta-inline" href="report.html">How to <i>leak</i> to us<span aria-hidden="true">↗</span></a>
+        </nav>
+      </header>
+
+      <!--
+        Edit this intro freely — it is static and will not be touched by
+        scripts-dev/build-references-page.js. Only the {{...}} placeholders
+        below are replaced by the build script.
+      -->
+      <section class="content-page__header-area" aria-labelledby="page-title">
+        <span class="content-page__kicker">References</span>
+        <h1 id="page-title">A working genealogy of face-recognition camouflage.</h1>
+        <p class="content-page__lead">
+          {{REFERENCE_COUNT}} references, {{YEAR_MIN}}&ndash;{{YEAR_MAX}}, newest first.
+        </p>
+        <p class="content-page__deck">
+          The Ghostmaxxing references archive maps artistic, academic, and activist work around face obfuscation,
+          adversarial makeup, physical-world attacks, and anti-surveillance aesthetics. Each entry documents what was
+          demonstrated, what its limits are, and whether Ghostmaxxing can directly or partially retest the idea.
+        </p>
+      </section>
+
+      <div class="content-page__layout">
+        <article class="content-page__main">
+          <section class="content-section" id="references-list">
+            <p class="content-page__lead reference-stats">{{STATS_LINE}}</p>
+
+            {{REFERENCES_BODY}}
+          </section>
+
+          <!-- Static section — edit freely -->
+          <section class="content-section" id="method">
+            <h2>How to suggest a new entry.</h2>
+            <p>
+              Every possible reference should be read and transformed into the format of <code>REFERENCES.json</code>,
+              check out on github
+              (https://github.com/vecna/ghostati/blob/rebranding/ghostmaxxing/references/REFERENCES.json) how detailed
+              is.
+              Once you're sure it might suit us, and potentially we can even program a Ghostyle to retest it, please
+              open an issue or submit a pull request.
+            </p>
+          </section>
+        </article>
+
+        <aside class="content-page__aside" aria-label="Page sections">
+          <h2>Jump to</h2>
+          <nav>
+            {{GROUP_NAV}}
+            <a href="#method">Suggest a new entry</a>
+          </nav>
+        </aside>
+      </div>
+    </div>
+
+    <div class="camera-band" aria-hidden="true"></div>
+  </main>
+</body>
+
+</html>
\ No newline at end of file
diff --git a/references/validate-references.js b/references/validate-references.js
new file mode 100644
index 0000000..b1160d5
--- /dev/null
+++ b/references/validate-references.js
@@ -0,0 +1,123 @@
+#!/usr/bin/env node
+'use strict';
+
+const fs = require('fs');
+const path = require('path');
+
+const ROOT = path.resolve(__dirname, '..');
+const REFERENCES_PATH = path.join(ROOT, 'references', 'REFERENCES.json');
+const TESTABILITY = new Set(['direct', 'partial', 'indirect', 'contextual', 'no']);
+const TYPES = new Set(['artistic', 'research', 'activism']);
+
+function readJson(filePath) {
+  return JSON.parse(fs.readFileSync(filePath, 'utf8'));
+}
+
+function isUrl(value) {
+  return typeof value === 'string' && /^https?:\/\//.test(value);
+}
+
+function isTag(value) {
+  return typeof value === 'string' && /^[a-z0-9-]+$/.test(value);
+}
+
+function assert(condition, message, errors) {
+  if (!condition) errors.push(message);
+}
+
+function validateReferences(data) {
+  const errors = [];
+  const warnings = [];
+
+  assert(data && typeof data === 'object', 'REFERENCES.json must contain an object.', errors);
+  assert(Array.isArray(data.references), 'references must be an array.', errors);
+  assert(data.language === 'en', 'language must be "en".', errors);
+
+  const tagDefs = data.tag_definitions || {};
+  const allowed = {
+    intervention: new Set(tagDefs.intervention || []),
+    target: new Set(tagDefs.target || []),
+    domain: new Set(tagDefs.domain || [])
+  };
+
+  const slugs = new Set();
+  const titles = new Set();
+  const refs = Array.isArray(data.references) ? data.references : [];
+
+  refs.forEach((ref, index) => {
+    const label = ref && ref.slug ? ref.slug : `entry #${index + 1}`;
+    assert(ref && typeof ref === 'object', `${label}: must be an object.`, errors);
+    if (!ref || typeof ref !== 'object') return;
+
+    assert(typeof ref.slug === 'string' && /^[a-z0-9]+(?:-[a-z0-9]+)*$/.test(ref.slug), `${label}: invalid slug.`, errors);
+    if (slugs.has(ref.slug)) errors.push(`${label}: duplicate slug.`);
+    slugs.add(ref.slug);
+
+    assert(typeof ref.title === 'string' && ref.title.trim().length > 0, `${label}: missing title.`, errors);
+    const titleKey = String(ref.title || '').trim().toLowerCase();
+    if (titleKey) {
+      if (titles.has(titleKey)) warnings.push(`${label}: duplicate normalized title detected: ${ref.title}`);
+      titles.add(titleKey);
+    }
+
+    assert(Array.isArray(ref.author) && ref.author.length > 0, `${label}: author must be a non-empty array.`, errors);
+    assert(Number.isInteger(ref.year) && ref.year >= 1900 && ref.year <= 2100, `${label}: year must be an integer.`, errors);
+    assert(isUrl(ref.link), `${label}: link must be an absolute http(s) URL.`, errors);
+    assert(TYPES.has(ref.type), `${label}: type must be artistic, research, or activism.`, errors);
+    assert(Number.isInteger(ref.closeness) && ref.closeness >= 1 && ref.closeness <= 100, `${label}: closeness must be an integer from 1 to 100.`, errors);
+
+    ['description', 'demonstrated', 'ghostati_relevance', 'limitations', 'reproducibility', 'citation'].forEach((field) => {
+      assert(typeof ref[field] === 'string' && ref[field].trim().length > 0, `${label}: missing ${field}.`, errors);
+    });
+
+    assert(ref.ghostati_testability && TESTABILITY.has(ref.ghostati_testability.level), `${label}: invalid ghostati_testability.level.`, errors);
+    assert(ref.ghostati_testability && typeof ref.ghostati_testability.reason === 'string' && ref.ghostati_testability.reason.trim().length > 0, `${label}: missing ghostati_testability.reason.`, errors);
+
+    ['intervention', 'target', 'domain'].forEach((field) => {
+      assert(Array.isArray(ref[field]) && ref[field].length > 0, `${label}: ${field} must be a non-empty array.`, errors);
+      (ref[field] || []).forEach((tag) => {
+        assert(isTag(tag), `${label}: ${field} contains invalid tag "${tag}".`, errors);
+        if (allowed[field].size && !allowed[field].has(tag)) {
+          errors.push(`${label}: ${field} tag "${tag}" is not declared in tag_definitions.${field}.`);
+        }
+      });
+    });
+
+    assert(ref.links && typeof ref.links === 'object', `${label}: links object is required.`, errors);
+    if (ref.links && typeof ref.links === 'object') {
+      ['canonical', 'paper', 'project', 'code', 'video', 'doi', 'arxiv'].forEach((field) => {
+        assert(Object.prototype.hasOwnProperty.call(ref.links, field), `${label}: links.${field} is required, use null if unknown.`, errors);
+        const value = ref.links[field];
+        assert(value === null || isUrl(value), `${label}: links.${field} must be null or an absolute http(s) URL.`, errors);
+      });
+    }
+  });
+
+  const newest = [...refs].sort((a, b) => (b.year - a.year) || (b.closeness - a.closeness) || String(a.title).localeCompare(String(b.title)));
+  refs.forEach((ref, index) => {
+    if (newest[index] && newest[index].slug !== ref.slug) {
+      warnings.push('references are not in default newest-first order; build output will sort them.');
+      return;
+    }
+  });
+
+  return { errors, warnings };
+}
+
+function main() {
+  const data = readJson(REFERENCES_PATH);
+  const { errors, warnings } = validateReferences(data);
+
+  warnings.forEach((warning) => console.warn(`warning: ${warning}`));
+
+  if (errors.length) {
+    errors.forEach((error) => console.error(`error: ${error}`));
+    process.exit(1);
+  }
+
+  console.log(`REFERENCES.json OK (${data.references.length} references).`);
+}
+
+if (require.main === module) main();
+
+module.exports = { validateReferences };
diff --git a/report.html b/report.html
new file mode 100644
index 0000000..1e802e2
--- /dev/null
+++ b/report.html
@@ -0,0 +1,264 @@
+<!doctype html>
+<html lang="en">
+
+<head>
+  <meta charset="utf-8" />
+  <meta name="viewport" content="width=device-width, initial-scale=1" />
+  <title>Leak to us — Ghostmaxxing</title>
+  <meta name="description"
+    content="A secure, anonymous reporting channel for anyone with direct or indirect knowledge of facial-recognition deployments — engineers, operators, procurement staff, bystanders, and people affected by it." />
+  <link rel="stylesheet" href="/styles/pages.css" />
+  <link rel="stylesheet" href="/styles/content-pages.css" />
+  <link rel="stylesheet" href="/styles/report-page.css" />
+</head>
+
+<body>
+  <main class="editorial-homepage content-page editorial-homepage--svg-footer-band">
+    <div class="editorial-homepage__shell">
+      <header class="editorial-homepage__header">
+        <a class="wordmark" href="index.html" aria-label="Ghostmaxxing homepage">
+          <span class="wordmark__title">Ghostmaxxing</span>
+          <span class="wordmark__subtitle">
+            Privacy activism meets public-research platform
+          </span>
+        </a>
+
+        <nav class="editorial-homepage__nav" aria-label="Project navigation">
+          <a class="navlink" href="https://github.com/vecna/ghostati" target="_blank" rel="noopener noreferrer">Code</a>
+          <a class="navlink" href="about.html">Vision &amp; About</a>
+          <a class="navlink" href="references.html">References</a>
+          <a class="cta-inline" href="report.html" aria-current="page">How to <i>leak</i> to us<span
+              aria-hidden="true">↗</span></a>
+        </nav>
+      </header>
+
+      <section class="content-page__header-area" aria-labelledby="page-title">
+        <span class="content-page__kicker">Leak to us</span>
+        <h1 id="page-title">To help us, tell us what you know.</h1>
+        <p class="content-page__lead">
+          Every deployment has people who understand it from the inside: engineers, operators, procurement staff,
+          bystanders, and the people it was pointed at. Whichever one you are, there's likely a specific set of
+          questions waiting for you below.
+        </p>
+        <p class="content-page__deck">
+          This page explains what to think about before you write anything, how the submission itself works, and
+          what happens after you send it. If you're wondering why we collect this at all, read
+          <a href="about.html#leaks">why leaking matters</a> first. Everything here runs on
+          <a href="https://www.globaleaks.org" target="_blank" rel="noopener noreferrer">GlobaLeaks</a>, the same
+          open-source system used by whistleblowing and journalism projects worldwide.
+        </p>
+      </section>
+
+      <div class="content-page__layout">
+        <article class="content-page__main">
+
+          <section class="content-section" id="questionnaires">
+            <h2>What we look for.</h2>
+            <p>
+              Each card below opens a separate, dedicated questionnaire we configured; Find if you fit in any of it, and
+              then keep reading the page for all the necessary security practices.
+            </p>
+            <div class="audience-grid">
+
+              <!-- TODO: replace each href below with
+                   https://raccontaci.nina.watch/#/submission?context=<uuid>
+                   once the matching context exists in the GlobaLeaks admin panel.
+                   Field-by-field specs for all eight questionnaires are in
+                   GLOBALEAKS-QUESTIONNAIRES.md. -->
+
+              <a class="audience-card" href="https://raccontaci.nina.watch" target="_blank" rel="noopener noreferrer">
+                <span class="audience-card__icon"><img src="/images/report/system-overview.svg" alt="" /></span>
+                <h3>You work with facial-recognition technology.</h3>
+                <p>You know how a system is built, what it covers, and who touches the data. Not every question
+                  needs an answer.</p>
+                <span class="audience-card__cta">Open this questionnaire</span>
+              </a>
+              <!--
+              <a class="audience-card" href="https://raccontaci.nina.watch" target="_blank" rel="noopener noreferrer">
+                <span class="audience-card__icon"><img src="/images/report/test-countermeasures.svg" alt="" /></span>
+                <h3>You can run our test cases against a real system.</h3>
+                <p>You have access to a live or staging deployment. We've this bundle of experiment to test, we would like to know if 
+                  you'll be run your tool over our samples, to know how the makeup is performing.</p>
+                <span class="audience-card__cta">Open this questionnaire</span>
+              </a>
+              COMMENTED BECAUSE WE DON'T HAVE THE BUNDLE YET
+-->
+
+              <a class="audience-card" href="https://raccontaci.nina.watch" target="_blank" rel="noopener noreferrer">
+                <span class="audience-card__icon"><img src="/images/report/weak-spots.svg" alt="" /></span>
+                <h3>You know where a deployment falls apart.</h3>
+                <p>You've seen where a specific system is misconfigured, poorly maintained, or easy to spot from
+                  the outside. Tell us what to look for, and where.</p>
+                <span class="audience-card__cta">Open this questionnaire</span>
+              </a>
+
+              <a class="audience-card" href="https://raccontaci.nina.watch" target="_blank" rel="noopener noreferrer">
+                <span class="audience-card__icon"><img src="/images/report/contracts-market.svg" alt="" /></span>
+                <h3>You've seen the paperwork.</h3>
+                <p>Contracts, tenders, vendor pitches, internal budgets. The commercial side of facial recognition
+                  is usually better hidden than the cameras themselves.</p>
+                <span class="audience-card__cta">Open this questionnaire</span>
+              </a>
+
+              <a class="audience-card" href="https://raccontaci.nina.watch" target="_blank" rel="noopener noreferrer">
+                <span class="audience-card__icon"><img src="/images/report/control-panel.svg" alt="" /></span>
+                <h3>You run the system, day to day.</h3>
+                <p>You're the one watching the alerts, overriding false matches, or logging who gets flagged. You
+                  know things the spec sheet doesn't say.</p>
+                <span class="audience-card__cta">Open this questionnaire</span>
+              </a>
+
+              <a class="audience-card" href="https://raccontaci.nina.watch" target="_blank" rel="noopener noreferrer">
+                <span class="audience-card__icon"><img src="/images/report/procurement-officials.svg" alt="" /></span>
+                <h3>You approved it, funded it, or signed off on it.</h3>
+                <p>Public tenders, impact assessments, oversight reports. Sitting on the buying side of the table
+                  means you saw a different version of the story.</p>
+                <span class="audience-card__cta">Open this questionnaire</span>
+              </a>
+
+              <a class="audience-card" href="https://raccontaci.nina.watch" target="_blank" rel="noopener noreferrer">
+                <span class="audience-card__icon"><img src="/images/report/public-sightings.svg" alt="" /></span>
+                <h3>You spotted something in public space.</h3>
+                <p>A camera, a kiosk, a sign, a badge you didn't recognize. You don't need inside access, but just a
+                  location and what you saw.</p>
+                <span class="audience-card__cta">Open this questionnaire</span>
+              </a>
+
+              <a class="audience-card" href="https://raccontaci.nina.watch" target="_blank" rel="noopener noreferrer">
+                <span class="audience-card__icon"><img src="/images/report/affected-individuals.svg" alt="" /></span>
+                <h3>You were stopped, flagged, or misidentified.</h3>
+                <p>Something went wrong on the other side of the lens, and it happened to you. Your experience is
+                  evidence too.</p>
+                <span class="audience-card__cta">Open this questionnaire</span>
+              </a>
+
+            </div>
+          </section>
+
+          <section class="content-section" id="before-you-write">
+            <h2>Before you write anything.</h2>
+            <p>
+              Most whistleblowers aren't exposed by a broken encryption algorithm. They're exposed by something
+              that happened before any tool was involved. A comment to a colleague, a search from a work laptop, a
+              detail that only one person could have known. None of what follows is about a specific piece of
+              software. It's about the decisions that come first.
+            </p>
+            <ul class="opsec-list">
+              <li>Don't mention your plan to colleagues, your manager, or friends who might repeat it without
+                realizing what it means.</li>
+              <li>Don't use your employer's device, account, or network to research this page, or to prepare
+                anything.</li>
+              <li>Photos and documents carry hidden information: timestamps, device IDs, GPS coordinates, printer
+                tracking dots.<b>We do not plan to release ANY of the material received</b>, but rather to use that as
+                investigation lead.
+                </br>
+                When in doubt, describe what you saw in your own words instead of uploading the
+                original file.</li>
+              <li>When you're ready, use the <a href="https://www.torproject.org/download/" target="_blank"
+                  rel="noopener noreferrer">Tor Browser</a> to open the submission page, ideally from a device and
+                network that isn't tied to your workplace.</li>
+            </ul>
+            <div class="callout">
+              <p>
+                These are practices we ask to our sources to protect themselves. Less secure channels might work too,
+                but we should assume to be surveilled even if our
+                research in the public interest is meant to strenghen the state of rights. so, please, follow the above
+                instructions, and if you need help, meet friendly digital rights organizations where you can get
+                additional support!
+              </p>
+            </div>
+
+            <h3>Know more, if you want to.</h3>
+            <p>
+              None of this reading is required before you submit anything. It's here for people who want to go
+              deeper on their own digital security, or who are doing this as part of an ongoing risk they're
+              managing, not a one-off report.
+            </p>
+            <div class="resource-grid">
+              <a class="resource-card" href="https://freedom.press/digisec/guides/source-protection/" target="_blank"
+                rel="noopener noreferrer">
+                <h3>Source protection guides</h3>
+                <p>Freedom of the Press Foundation on reaching a journalist safely, written for people on your side
+                  of the exchange.</p>
+              </a>
+              <a class="resource-card" href="https://ssd.eff.org/" target="_blank" rel="noopener noreferrer">
+                <h3>Surveillance Self-Defense</h3>
+                <p>EFF's general guide to digital security, from the basics to specific tools and scenarios.</p>
+              </a>
+              <a class="resource-card" href="https://holistic-security.tacticaltech.org/" target="_blank"
+                rel="noopener noreferrer">
+                <h3>Holistic Security</h3>
+                <p>Tactical Tech's guide for activists — treats digital, physical, and psychological safety as one
+                  problem.</p>
+              </a>
+              <a class="resource-card" href="https://securityinabox.org/" target="_blank" rel="noopener noreferrer">
+                <h3>Security in a Box</h3>
+                <p>Step-by-step digital security guides organized by device and by task.</p>
+              </a>
+            </div>
+          </section>
+
+          <section class="content-section" id="how-it-works">
+            <h2>How the submission works.</h2>
+            <p>
+              GlobaLeaks organizes each reporting channel into a <strong>questionnaire</strong> — a set of
+              questions tailored to a specific kind of knowledge. Below, you'll pick the one that matches what you
+              actually know, rather than filling in one generic form built for no one in particular.
+            </p>
+            <ul class="opsec-list">
+              <li>Most fields are optional. Answer what you can and skip the rest — a partial answer is still
+                useful to us.</li>
+              <li>You don't create an account, and nothing in the form asks for your name or contact details.</li>
+              <li>The submission itself happens inside GlobaLeaks, reachable over Tor, the same system used by
+                whistleblowing projects at newsrooms and NGOs worldwide.</li>
+            </ul>
+          </section>
+
+          <section class="content-section" id="after-you-submit">
+            <h2>Submitting is the beginning, not the end.</h2>
+            <p>
+              A real person on our editorial and research team reads every submission — not an automated filter.
+              Before you send anything, the submission screen shows you who's set to receive it, so you're not
+              dropping this into an anonymous inbox on our side, even while you stay anonymous on yours.
+            </p>
+            <p>
+              At the end of the form, you'll be given a private <strong>receipt code</strong>.
+            </p>
+            <div class="receipt-callout">
+              <span class="receipt-callout__code" aria-hidden="true">•••• •••• •••• ••••</span>
+              <p>
+                That code is the only way — for you or for us — to reopen your case. Save it somewhere safe,
+                offline if you can. If it's lost, the conversation can't be recovered. That's by design, not a bug.
+              </p>
+            </div>
+            <p>
+              Come back and use it. We may have follow-up questions, and a source who keeps that thread open tends
+              to help an investigation go much further than a single anonymous drop ever could. Secure reporting
+              works best as part of a wider network of technical, legal, and editorial support around it, not as a
+              black box that swallows a report and gives nothing back — that's the same thinking behind efforts like
+              GlobaLeaks' own <a
+                href="https://globaleaks.org/2026/06/23/supporting-the-house-of-whistleblowers-expanding-the-ecosystem-around-secure-reporting/"
+                target="_blank" rel="noopener noreferrer">House of Whistleblowers</a> initiative, and it's why we
+              treat the code above as the start of a conversation, not a drop box receipt.
+            </p>
+          </section>
+
+        </article>
+
+        <aside class="content-page__aside" aria-label="Page sections">
+          <h2>Jump to</h2>
+          <nav>
+            <a href="#questionnaires">What we look for</a>
+            <a href="#before-you-write">Before you write anything</a>
+            <a href="#how-it-works">How the submission works</a>
+            <a href="#after-you-submit">After you submit</a>
+          </nav>
+        </aside>
+      </div>
+    </div>
+
+    <div class="camera-band" aria-hidden="true"></div>
+  </main>
+</body>
+
+</html>
\ No newline at end of file
diff --git a/scripts-dev/validate-plugin.js b/scripts-dev/validate-plugin.js
old mode 100644
new mode 100755
diff --git a/scripts/Ghostati.d.ts b/scripts/Ghostati.d.ts
index 4d00ec0..b6bdeba 100644
--- a/scripts/Ghostati.d.ts
+++ b/scripts/Ghostati.d.ts
@@ -1,5 +1,5 @@
 /**
- * Ghostati public API declarations for JS plugin authors.
+ * Ghostmaxxing (Ghostati) public API declarations for JS plugin authors.
  *
  * This file is intentionally declaration-only (no runtime side effects).
  * Editors like VS Code pick it up automatically to provide autocomplete
@@ -251,6 +251,8 @@ export interface GhostatiApi {
 declare global {
   interface Window {
     Ghostati: GhostatiApi;
+    /** International alias – identical to {@link GhostatiApi Ghostati}. */
+    Ghostmaxxing: GhostatiApi;
   }
 }
 
diff --git a/scripts/analyze-panel.js b/scripts/analyze-panel.js
index 111d88a..c32f598 100644
--- a/scripts/analyze-panel.js
+++ b/scripts/analyze-panel.js
@@ -396,7 +396,7 @@ export function generateReportText() {
    const data = latestReportData;
    if (!data) {
       return [
-         '### Ghostati - Analisi del trucco',
+         '### Ghostmaxxing - Analisi del trucco',
          '',
          'Volto rilevato: no',
          '',
@@ -407,7 +407,7 @@ export function generateReportText() {
    }
 
    const lines = [];
-   lines.push('### Ghostati - Analisi del trucco');
+   lines.push('### Ghostmaxxing - Analisi del trucco');
    lines.push('');
    lines.push(`Volto rilevato: ${data.faceResult ? 'si' : 'no'}`);
 
diff --git a/scripts/config.js b/scripts/config.js
index 658be7a..b8f51bc 100644
--- a/scripts/config.js
+++ b/scripts/config.js
@@ -1,7 +1,7 @@
 /**
  * @module config
  * @description
- * Centralised configuration constants for Ghostati. Lives in one place so that
+ * Centralised configuration constants for Ghostmaxxing. Lives in one place so that
  * URLs, detector thresholds, and recording defaults can be changed without
  * touching engine code. All values here are static and module-load-time
  * resolvable; runtime-mutable state belongs in `state.js`.
diff --git a/scripts/export-makeup.js b/scripts/export-makeup.js
index 1b88323..d3973b5 100644
--- a/scripts/export-makeup.js
+++ b/scripts/export-makeup.js
@@ -14,7 +14,7 @@ const EXPORT_LAYOUT = {
 const EXPORT_COPY = {
    filename: 'ghostati-makeup.png',
    mimeType: 'image/png',
-   shareTitle: 'Ghostati Makeup',
+   shareTitle: 'Ghostmaxxing Makeup',
    shareText: 'Il mio camouflage anti-riconoscimento!',
    headerPrefix: 'github.com/vecna/ghostati | Report diagnostico makeup',
    reportUrl: 'https://vecna.eu/ghostmaxxing//'
diff --git a/scripts/ghostati-mobile-ui.js b/scripts/ghostati-mobile-ui.js
index abfdb7c..6c00b07 100644
--- a/scripts/ghostati-mobile-ui.js
+++ b/scripts/ghostati-mobile-ui.js
@@ -61,6 +61,11 @@ document.addEventListener('DOMContentLoaded', () => {
     return el.closest('.scrollable') || el.closest('#settingsDrawer') || el.closest('#historyDrawer');
   };
 
+  /* [SYSTEM API: document.addEventListener('touchstart'/'touchend')]
+   * Funzionamento: API nativa per intercettare il tocco delle dita sui display touch.
+   * Parametri: 'touchstart', callback(e), { passive: true } (migliora le performance dicendo al browser che non useremo preventDefault).
+   * Feature: Registra l'inizio e la fine di uno swipe verticale per attivare la pulizia dei log e dell'overlay (gesture UI).
+   */
   document.addEventListener('touchstart', (e) => {
     if (isScrollableElement(e.target)) return;
     touchStartY = e.changedTouches[0].screenY;
@@ -72,7 +77,11 @@ document.addEventListener('DOMContentLoaded', () => {
     handleGesture();
   }, { passive: true });
 
-  // Scroll wheel for desktop
+  /* [SYSTEM API: document.addEventListener('wheel')]
+   * Funzionamento: Intercetta la rotellina del mouse o il trackpad su desktop.
+   * Parametri: 'wheel', callback(e), { passive: true }.
+   * Feature: Replica il comportamento dello swipe (clear overlay e log) anche per gli utenti desktop tramite wheel/scroll.
+   */
   document.addEventListener('wheel', (e) => {
     if (isScrollableElement(e.target)) return;
     // Debounce or just trigger on any significant scroll
diff --git a/scripts/index-effect.js b/scripts/index-effect.js
deleted file mode 100644
index 43d1ef0..0000000
--- a/scripts/index-effect.js
+++ /dev/null
@@ -1,298 +0,0 @@
-/** @module index-effect */
-const hero = document.getElementById('ghostHero');
-const link = document.getElementById('ghostLink');
-const veil = document.getElementById('ghostVeil');
-const topbar = document.getElementById('ghostTopbar');
-const progressDisplay = document.getElementById('revealProgress');
-const menuToggle = document.getElementById('ghostMenuToggle');
-const topbarMenu = document.getElementById('ghostTopbarMenu');
-
-if (!hero || !link || !veil) {
-  throw new Error('Ghostati hero elements not found.');
-}
-
-const context = veil.getContext('2d', { alpha: true });
-if (!context) {
-  throw new Error('Canvas 2D context is not available.');
-}
-
-const effectState = {
-  unlocked: false,
-  revealRatio: 0,
-  pointerX: 0,
-  pointerY: 0,
-  hasPointer: false,
-  interactionQueued: false,
-  resizeQueued: false,
-  unlockBlinkTimeout: null,
-};
-
-const SETTINGS = {
-  brushRadius: 40,
-  brushEdge: 70,
-  unlockThreshold: 0.34,
-  sampleStep: 18,
-};
-
-function clamp(value, min = 0, max = 1) {
-  return Math.min(max, Math.max(min, value));
-}
-
-function updateUnlockState() {
-  const shouldUnlock = effectState.revealRatio >= SETTINGS.unlockThreshold;
-  if (shouldUnlock === effectState.unlocked) {
-    return;
-  }
-
-  effectState.unlocked = shouldUnlock;
-  hero.classList.toggle('is-unlocked', shouldUnlock);
-  topbar?.classList.toggle('is-unlocked', shouldUnlock);
-  link.setAttribute('aria-disabled', String(!shouldUnlock));
-  link.tabIndex = shouldUnlock ? 0 : -1;
-
-  if (effectState.unlockBlinkTimeout) {
-    clearTimeout(effectState.unlockBlinkTimeout);
-    effectState.unlockBlinkTimeout = null;
-  }
-
-  if (shouldUnlock) {
-    hero.classList.add('just-unlocked');
-    effectState.unlockBlinkTimeout = window.setTimeout(() => {
-      hero.classList.remove('just-unlocked');
-      effectState.unlockBlinkTimeout = null;
-    }, 1400);
-    return;
-  }
-
-  hero.classList.remove('just-unlocked');
-}
-
-function updateProgressDisplay() {
-  if (!progressDisplay) {
-    return;
-  }
-
-  const percentage = Math.round(clamp(effectState.revealRatio) * 100);
-  progressDisplay.textContent = `${percentage}%`;
-}
-
-function fillVeil() {
-  context.globalCompositeOperation = 'source-over';
-  context.fillStyle = '#000000';
-  context.fillRect(0, 0, veil.width, veil.height);
-}
-
-function indexResize(preserveReveal = true) {
-  console.log("check: quando è usato? serve davvero? è uguale a camera.resizeCanvas?")
-  const bounds = hero.getBoundingClientRect();
-  const width = Math.max(1, Math.floor(bounds.width));
-  const height = Math.max(1, Math.floor(bounds.height));
-
-  const previous = preserveReveal
-    ? context.getImageData(0, 0, veil.width || 1, veil.height || 1)
-    : null;
-
-  veil.width = width;
-  veil.height = height;
-
-  fillVeil();
-
-  if (previous) {
-    const copyCanvas = document.createElement('canvas');
-    copyCanvas.width = previous.width;
-    copyCanvas.height = previous.height;
-    const copyCtx = copyCanvas.getContext('2d');
-
-    if (copyCtx) {
-      copyCtx.putImageData(previous, 0, 0);
-      context.globalCompositeOperation = 'destination-out';
-      context.drawImage(copyCanvas, 0, 0, width, height);
-      context.globalCompositeOperation = 'source-over';
-    }
-  }
-
-  updateRevealRatio();
-}
-
-function eraseAt(clientX, clientY) {
-  const rect = veil.getBoundingClientRect();
-  const x = clientX - rect.left;
-  const y = clientY - rect.top;
-
-  const innerRadius = SETTINGS.brushRadius;
-  const outerRadius = SETTINGS.brushRadius + SETTINGS.brushEdge;
-
-  const gradient = context.createRadialGradient(x, y, innerRadius, x, y, outerRadius);
-  gradient.addColorStop(0, 'rgba(0, 0, 0, 1)');
-  gradient.addColorStop(1, 'rgba(0, 0, 0, 0)');
-
-  context.globalCompositeOperation = 'destination-out';
-  context.fillStyle = gradient;
-  context.beginPath();
-  context.arc(x, y, outerRadius, 0, Math.PI * 2);
-  context.fill();
-  context.globalCompositeOperation = 'source-over';
-}
-
-function updateRevealRatio() {
-  const width = veil.width;
-  const height = veil.height;
-  const step = SETTINGS.sampleStep;
-
-  let sampled = 0;
-  let transparent = 0;
-
-  for (let y = 0; y < height; y += step) {
-    const sampleHeight = Math.min(step, height - y);
-    const row = context.getImageData(0, y, width, sampleHeight).data;
-
-    for (let i = 3; i < row.length; i += 4 * step) {
-      sampled += 1;
-      if (row[i] < 220) {
-        transparent += 1;
-      }
-    }
-  }
-
-  effectState.revealRatio = sampled ? clamp(transparent / sampled) : 0;
-  updateProgressDisplay();
-  updateUnlockState();
-}
-
-function runInteractionFrame() {
-  effectState.interactionQueued = false;
-  if (!effectState.hasPointer) {
-    return;
-  }
-
-  eraseAt(effectState.pointerX, effectState.pointerY);
-  updateRevealRatio();
-}
-
-function queueInteractionFrame() {
-  if (effectState.interactionQueued) {
-    return;
-  }
-
-  effectState.interactionQueued = true;
-  requestAnimationFrame(runInteractionFrame);
-}
-
-function queueResize() {
-  if (effectState.resizeQueued) {
-    return;
-  }
-
-  effectState.resizeQueued = true;
-  requestAnimationFrame(() => {
-    effectState.resizeQueued = false;
-    indexResize(true);
-  });
-}
-
-function handlePointerMove(clientX, clientY) {
-  effectState.pointerX = clientX;
-  effectState.pointerY = clientY;
-  effectState.hasPointer = true;
-  queueInteractionFrame();
-}
-
-function interactionTargetsTopbar(target) {
-  return target instanceof Element && Boolean(target.closest('#ghostTopbar'));
-}
-
-window.addEventListener('pointermove', (event) => {
-  if (interactionTargetsTopbar(event.target)) {
-    return;
-  }
-
-  handlePointerMove(event.clientX, event.clientY);
-});
-
-window.addEventListener('pointerdown', (event) => {
-  if (interactionTargetsTopbar(event.target)) {
-    return;
-  }
-
-  handlePointerMove(event.clientX, event.clientY);
-});
-
-window.addEventListener('touchstart', (event) => {
-  if (interactionTargetsTopbar(event.target)) {
-    return;
-  }
-
-  const touch = event.touches?.[0];
-  if (!touch) {
-    return;
-  }
-
-  handlePointerMove(touch.clientX, touch.clientY);
-}, { passive: true });
-
-window.addEventListener('touchmove', (event) => {
-  if (interactionTargetsTopbar(event.target)) {
-    return;
-  }
-
-  const touch = event.touches?.[0];
-  if (!touch) {
-    return;
-  }
-
-  handlePointerMove(touch.clientX, touch.clientY);
-}, { passive: true });
-
-window.addEventListener('resize', queueResize);
-
-window.addEventListener('resize', () => {
-  if (window.matchMedia('(min-width: 901px) and (pointer: fine)').matches) {
-    topbar?.classList.remove('is-open');
-    menuToggle?.setAttribute('aria-expanded', 'false');
-  }
-});
-
-window.addEventListener('keydown', (event) => {
-  if (event.key === 'Escape') {
-    topbar?.classList.remove('is-open');
-    menuToggle?.setAttribute('aria-expanded', 'false');
-  }
-
-  if (event.key === 'Enter' && !effectState.unlocked) {
-    event.preventDefault();
-  }
-});
-
-link.addEventListener('click', (event) => {
-  if (!effectState.unlocked) {
-    event.preventDefault();
-  }
-});
-
-menuToggle?.addEventListener('click', () => {
-  const nextOpen = !topbar?.classList.contains('is-open');
-  topbar?.classList.toggle('is-open', nextOpen);
-  menuToggle.setAttribute('aria-expanded', String(Boolean(nextOpen)));
-});
-
-topbarMenu?.querySelectorAll('a').forEach((anchor) => {
-  anchor.addEventListener('click', () => {
-    topbar?.classList.remove('is-open');
-    menuToggle?.setAttribute('aria-expanded', 'false');
-  });
-});
-
-window.addEventListener('pointerdown', (event) => {
-  const target = event.target;
-  const clickedOutsideTopbar = target instanceof Element && !target.closest('#ghostTopbar');
-  if (!clickedOutsideTopbar) {
-    return;
-  }
-
-  topbar?.classList.remove('is-open');
-  menuToggle?.setAttribute('aria-expanded', 'false');
-});
-
-indexResize(false);
-updateProgressDisplay();
-updateUnlockState();
diff --git a/scripts/main.js b/scripts/main.js
index dce1426..d086c55 100644
--- a/scripts/main.js
+++ b/scripts/main.js
@@ -107,6 +107,8 @@ window.Ghostati = {
    detectorOptions: DETECTOR_OPTIONS
 };
 
+// International alias (see redesign/ghostati-ghostmaxxing-naming-brief.txt)
+window.Ghostmaxxing = window.Ghostati;
 
 /**
  * Sets the busy flag for the whole UI, disabling/enabling controls during asynchronous operations.
diff --git a/scripts/mediapipe-loop.js b/scripts/mediapipe-loop.js
index e127c93..afdc12c 100644
--- a/scripts/mediapipe-loop.js
+++ b/scripts/mediapipe-loop.js
@@ -87,7 +87,7 @@ function tick() {
       const results = faceLandmarker.detectForVideo(video, now);
       const landmarks = (results.faceLandmarks && results.faceLandmarks[0]) || null;
 
-      // Cache latest landmarks on state (via the Ghostati setter) so engine-3d
+      // Cache latest landmarks on state (via the Ghostmaxxing setter) so engine-3d
       // and plugins can read them without re-running inference.
       if (window.Ghostati) window.Ghostati.lastLandmarks3d = landmarks;
 
diff --git a/scripts/state.js b/scripts/state.js
index d62619f..781e9f6 100644
--- a/scripts/state.js
+++ b/scripts/state.js
@@ -1,12 +1,12 @@
 /**
  * @module state
  * @description
- * Single source of mutable runtime state for the Ghostati web app. A plain
+ * Single source of mutable runtime state for the Ghostmaxxing web app. A plain
  * object shared (by reference) across all ES modules via `import { state }`.
  * There is exactly one instance; mutating a field here is observed everywhere.
  *
  * The full shape — including the type, allowed range, and consuming modules of
- * every field — is documented on the {@link GhostatiState} typedef below, which
+ * every field — is documented on the {@link GhostmaxxingState} typedef below, which
  * renders as a properties table.
  *
  * Note on events: the in-app event bus is `state.ghostatiEvents`. The
@@ -42,7 +42,7 @@
  * Shape of the shared runtime state object. Each property documents its type,
  * its allowed/observed range, and the modules that read or write it.
  *
- * @typedef {Object} GhostatiState
+ * @typedef {Object} GhostmaxxingState
  *
  * @property {{ nextId: number, faces: FaceRecord[] } | null} db
  *   Local face database (recognition descriptors + id counter).
@@ -191,7 +191,7 @@
 
 /**
  * The shared runtime state singleton.
- * @type {GhostatiState}
+ * @type {GhostmaxxingState}
  */
 export const state = {
    db: null, // initialized after loading via db.loadDb()
diff --git a/sitemap.xml b/sitemap.xml
index 9ea52aa..08c7005 100644
--- a/sitemap.xml
+++ b/sitemap.xml
@@ -1,15 +1,70 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
+   <!-- Core Ghòstati pages and project resources. -->
    <url>
       <loc>https://sindacato.nina.watch/ghostati/</loc>
-      <lastmod>2026-05-04</lastmod>
+      <lastmod>2026-06-29</lastmod>
       <changefreq>weekly</changefreq>
       <priority>1.0</priority>
    </url>
+   <url>
+      <loc>https://sindacato.nina.watch/ghostati/ghostati.html</loc>
+      <lastmod>2026-06-29</lastmod>
+      <changefreq>weekly</changefreq>
+      <priority>0.95</priority>
+   </url>
    <url>
       <loc>https://sindacato.nina.watch/ghostati/ghostati-docs.html</loc>
-      <lastmod>2026-04-13</lastmod>
+      <lastmod>2026-06-29</lastmod>
       <changefreq>monthly</changefreq>
       <priority>0.8</priority>
    </url>
+   <url>
+      <loc>https://sindacato.nina.watch/ghostati/docs/</loc>
+      <lastmod>2026-06-29</lastmod>
+      <changefreq>weekly</changefreq>
+      <priority>0.8</priority>
+   </url>
+   <url>
+      <loc>https://sindacato.nina.watch/ghostati/coverage/</loc>
+      <lastmod>2026-06-29</lastmod>
+      <changefreq>weekly</changefreq>
+      <priority>0.6</priority>
+   </url>
+   <url>
+      <loc>https://sindacato.nina.watch/ghostati/REFERENCES.json</loc>
+      <lastmod>2026-06-29</lastmod>
+      <changefreq>monthly</changefreq>
+      <priority>0.7</priority>
+   </url>
+   <url>
+      <loc>https://sindacato.nina.watch/ghostati/PROMPT-REFERENCES-UPDATE.txt</loc>
+      <lastmod>2026-06-29</lastmod>
+      <changefreq>monthly</changefreq>
+      <priority>0.5</priority>
+   </url>
+   <url>
+      <loc>https://ghostyles.vecna.eu/</loc>
+      <lastmod>2026-06-29</lastmod>
+      <changefreq>weekly</changefreq>
+      <priority>0.75</priority>
+   </url>
+   <url>
+      <loc>https://github.com/vecna/ghostati</loc>
+      <lastmod>2026-06-29</lastmod>
+      <changefreq>weekly</changefreq>
+      <priority>0.65</priority>
+   </url>
+   <url>
+      <loc>https://sindacato.nina.watch/it/iniziative/ghostati/</loc>
+      <lastmod>2026-06-29</lastmod>
+      <changefreq>monthly</changefreq>
+      <priority>0.7</priority>
+   </url>
+   <url>
+      <loc>https://raccontaci.nina.watch/#/submission?context=10c78596-3ea0-4867-b2fb-21fdb8e3f40c</loc>
+      <lastmod>2026-06-29</lastmod>
+      <changefreq>weekly</changefreq>
+      <priority>0.9</priority>
+   </url>
 </urlset>
diff --git a/styles/content-pages.css b/styles/content-pages.css
new file mode 100644
index 0000000..036f510
--- /dev/null
+++ b/styles/content-pages.css
@@ -0,0 +1,275 @@
+/* Ghostmaxxing secondary pages.
+ * Include after `styles/pages.css`.
+ * This file defines reusable typography and content components for pages such as
+ * about.html, report.html, references intros, workshop pages, and strategy notes.
+ */
+
+.content-page {
+  min-height: 100svh;
+}
+
+.content-page .editorial-homepage__shell {
+  padding-bottom: calc(var(--camera-band-height) * 0.78 + 3.5rem);
+}
+
+.content-page__header-area {
+  max-width: 68rem;
+  padding-top: clamp(2.25rem, 5vw, 5rem);
+  padding-bottom: clamp(2rem, 4vw, 4rem);
+}
+
+.content-page__kicker {
+  display: inline-flex;
+  align-items: center;
+  width: fit-content;
+  margin-bottom: 1rem;
+  padding: 0.42rem 0.62rem;
+  border: 1px solid rgba(5, 5, 5, 0.24);
+  background: rgba(255, 231, 168, 0.12);
+  color: var(--gm-text);
+  font-size: 0.78rem;
+  font-weight: 780;
+  letter-spacing: 0.08em;
+  line-height: 1;
+  text-transform: uppercase;
+}
+
+.content-page h1,
+.content-page h2,
+.content-page h3 {
+  margin: 0;
+  font-family: var(--serif);
+  color: var(--gm-ink);
+  font-weight: 700;
+  letter-spacing: -0.045em;
+}
+
+.content-page h1 {
+  max-width: 14ch;
+  font-size: clamp(4.1rem, 10vw, 9.4rem);
+  line-height: 0.88;
+  letter-spacing: -0.065em;
+}
+
+.content-page h2 {
+  max-width: 18ch;
+  margin-top: clamp(2.5rem, 5vw, 5rem);
+  font-size: clamp(2.45rem, 4.6vw, 5.25rem);
+  line-height: 0.92;
+}
+
+.content-page h3 {
+  max-width: 34rem;
+  margin-top: 1.65rem;
+  font-size: clamp(1.55rem, 1.4vw + 1.05rem, 2.55rem);
+  line-height: 1;
+  letter-spacing: -0.032em;
+}
+
+.content-page p {
+  max-width: 45rem;
+  margin: 0;
+  color: var(--gm-text);
+  font-size: clamp(1.02rem, 0.25vw + 0.98rem, 1.18rem);
+  line-height: 1.54;
+}
+
+.content-page p + p {
+  margin-top: 1rem;
+}
+
+.content-page__lead {
+  max-width: 52rem !important;
+  margin-top: 1.35rem !important;
+  color: var(--gm-text) !important;
+  font-size: clamp(1.25rem, 1.2vw + 1rem, 2rem) !important;
+  line-height: 1.16 !important;
+  font-weight: 650;
+}
+
+.content-page__deck {
+  max-width: 48rem !important;
+  margin-top: 1.1rem !important;
+  color: var(--gm-muted) !important;
+  font-size: clamp(1rem, 0.45vw + 0.95rem, 1.22rem) !important;
+  line-height: 1.48 !important;
+}
+
+.content-page__layout {
+  display: grid;
+  grid-template-columns: minmax(0, 0.72fr) minmax(16rem, 0.28fr);
+  gap: clamp(2rem, 5vw, 5rem);
+  align-items: start;
+}
+
+.content-page__main {
+  min-width: 0;
+}
+
+.content-section {
+  padding-top: 2rem;
+  border-top: 1px solid rgba(7, 26, 51, 0.22);
+}
+
+.content-section:first-child {
+  border-top: 0;
+  padding-top: 0;
+}
+
+.content-section p:first-of-type {
+  margin-top: 1rem;
+}
+
+.content-section ul,
+.content-section ol {
+  max-width: 45rem;
+  margin: 1rem 0 0 1.2rem;
+  padding: 0;
+  color: var(--gm-text);
+  font-size: clamp(1.02rem, 0.25vw + 0.98rem, 1.18rem);
+  line-height: 1.5;
+}
+
+.content-section li + li {
+  margin-top: 0.55rem;
+}
+
+.content-page strong {
+  color: var(--gm-ink);
+  font-weight: 800;
+}
+
+.content-page a:not(.wordmark):not(.navlink):not(.cta-inline):not(.btn-primary):not(.link-secondary) {
+  color: var(--gm-ink);
+  font-weight: 760;
+  text-decoration: none;
+  padding-inline: 0.08em;
+  background:
+    linear-gradient(180deg,
+      transparent 56%,
+      rgba(243, 199, 71, 0.72) 56%,
+      rgba(243, 199, 71, 0.72) 90%,
+      transparent 90%);
+}
+
+.content-page a:not(.wordmark):not(.navlink):not(.cta-inline):not(.btn-primary):not(.link-secondary):hover,
+.content-page a:not(.wordmark):not(.navlink):not(.cta-inline):not(.btn-primary):not(.link-secondary):focus-visible {
+  background: var(--gm-cream);
+  color: var(--gm-ink);
+  outline: 2px solid rgba(5, 5, 5, 0.55);
+  outline-offset: 0.12em;
+}
+
+.content-page__aside {
+  position: sticky;
+  top: 1.5rem;
+  padding: 1rem;
+  border: 1px solid rgba(7, 26, 51, 0.28);
+  background: rgba(255, 231, 168, 0.13);
+  box-shadow: 5px 5px 0 rgba(5, 5, 5, 0.12);
+}
+
+.content-page__aside h2 {
+  margin: 0 0 0.85rem;
+  max-width: none;
+  font-family: var(--sans);
+  font-size: 0.78rem;
+  line-height: 1;
+  letter-spacing: 0.08em;
+  text-transform: uppercase;
+}
+
+.content-page__aside nav {
+  display: grid;
+  gap: 0.55rem;
+}
+
+.content-page__aside a {
+  width: fit-content;
+  font-size: 0.96rem;
+}
+
+.callout {
+  max-width: 52rem;
+  margin-top: 1.4rem;
+  padding: 1rem 1rem 1rem 1.2rem;
+  border-left: 5px solid var(--gm-text);
+  background: rgba(255, 231, 168, 0.16);
+  color: var(--gm-text);
+}
+
+.callout p {
+  max-width: none;
+  font-weight: 650;
+}
+
+.resource-grid {
+  display: grid;
+  grid-template-columns: repeat(2, minmax(0, 1fr));
+  gap: 1rem;
+  max-width: 58rem;
+  margin-top: 1.25rem;
+}
+
+.resource-card {
+  display: block;
+  min-height: 9rem;
+  padding: 1rem;
+  border: 1px solid rgba(7, 26, 51, 0.32);
+  background: rgba(255, 231, 168, 0.13);
+  box-shadow: 4px 4px 0 rgba(5, 5, 5, 0.13);
+}
+
+.resource-card h3 {
+  margin: 0;
+  font-family: var(--sans);
+  font-size: 0.78rem;
+  letter-spacing: 0.08em;
+  line-height: 1.1;
+  text-transform: uppercase;
+}
+
+.resource-card p {
+  margin-top: 0.65rem;
+  color: var(--gm-muted);
+  font-size: 0.96rem;
+  line-height: 1.42;
+}
+
+.content-page__actions {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 1rem;
+  margin-top: 1.5rem;
+}
+
+@media (max-width: 980px) {
+  .content-page__layout {
+    grid-template-columns: 1fr;
+  }
+
+  .content-page__aside {
+    position: static;
+    order: -1;
+  }
+
+  .content-page__aside nav {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 0.7rem 1rem;
+  }
+}
+
+@media (max-width: 680px) {
+  .content-page h1 {
+    font-size: clamp(3.6rem, 18vw, 5.4rem);
+  }
+
+  .content-page h2 {
+    font-size: clamp(2.1rem, 12vw, 3.7rem);
+  }
+
+  .resource-grid {
+    grid-template-columns: 1fr;
+  }
+}
diff --git a/styles/ghostati-docs.css b/styles/ghostati-docs.css
deleted file mode 100644
index 64c114e..0000000
--- a/styles/ghostati-docs.css
+++ /dev/null
@@ -1,136 +0,0 @@
-:root {
-  --bg: #0f1115;
-  --panel: #171a21;
-  --panel-2: #1d2230;
-  --text: #eef2ff;
-  --muted: #aeb6cf;
-  --line: #2c3346;
-  --accent: #7aa2ff;
-  --accent-2: #9f7aea;
-  --success: #3ddc97;
-  --warn: #ffcc66;
-  --danger: #ff7a7a;
-  --shadow: 0 16px 40px rgba(0, 0, 0, 0.28);
-  --radius: 18px;
-}
-
-* { box-sizing: border-box; }
-html, body { height: 100%; }
-body {
-  margin: 0;
-  background:
-    radial-gradient(circle at top left, rgba(122, 162, 255, 0.14), transparent 28%),
-    radial-gradient(circle at top right, rgba(159, 122, 234, 0.12), transparent 26%),
-    var(--bg);
-  color: var(--text);
-  font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif;
-  line-height: 1.6;
-}
-
-.page {
-  max-width: 900px;
-  margin: 0 auto;
-  padding: 40px 24px;
-  display: flex;
-  flex-direction: column;
-  gap: 24px;
-}
-
-header {
-  background: linear-gradient(135deg, rgba(122, 162, 255, 0.16), rgba(159, 122, 234, 0.12));
-  border: 1px solid var(--line);
-  border-radius: var(--radius);
-  box-shadow: var(--shadow);
-  padding: 30px;
-  text-align: center;
-}
-
-h1 {
-  margin: 0 0 10px;
-  font-size: 32px;
-  font-weight: 800;
-  letter-spacing: 0.02em;
-}
-
-.subtitle {
-  color: var(--muted);
-  font-size: 16px;
-}
-
-.btn-back {
-  display: inline-block;
-  margin-top: 20px;
-  text-decoration: none;
-  background: linear-gradient(180deg, rgba(255, 255, 255, 0.08), rgba(255, 255, 255, 0.03));
-  border: 1px solid rgba(255, 255, 255, 0.06);
-  color: var(--muted);
-  border-radius: 14px;
-  font-size: 14px;
-  font-weight: 600;
-  padding: 10px 20px;
-  transition: all 0.2s ease;
-}
-
-.btn-back:hover {
-  background: linear-gradient(180deg, rgba(255, 255, 255, 0.12), rgba(255, 255, 255, 0.06));
-  color: var(--text);
-  transform: translateY(-1px);
-}
-
-section {
-  background: linear-gradient(180deg, rgba(255, 255, 255, 0.015), rgba(255, 255, 255, 0));
-  border: 1px solid var(--line);
-  border-radius: var(--radius);
-  box-shadow: var(--shadow);
-  padding: 32px;
-}
-
-h2 {
-  margin: 0 0 16px;
-  font-size: 24px;
-  font-weight: 700;
-  color: var(--accent);
-  border-bottom: 1px solid var(--line);
-  padding-bottom: 8px;
-}
-
-h3 {
-  font-size: 18px;
-  color: var(--accent-2);
-  margin-top: 30px;
-}
-
-p {
-  color: var(--muted);
-  margin-bottom: 16px;
-}
-
-pre {
-  background: #0a0d12;
-  border: 1px solid var(--line);
-  border-radius: 12px;
-  padding: 20px;
-  overflow-x: auto;
-  font-size: 14px;
-  line-height: 1.5;
-}
-
-code {
-  font-family: ui-monospace, SFMono-Regular, Menlo, Monaco, Consolas, fallback;
-  color: #c6d0f5;
-}
-
-.block-explain {
-  background: rgba(255, 255, 255, 0.03);
-  border-left: 4px solid var(--accent);
-  padding: 16px 20px;
-  margin-bottom: 16px;
-  border-radius: 0 12px 12px 0;
-}
-
-.block-explain strong {
-  display: block;
-  color: var(--text);
-  margin-bottom: 8px;
-  font-size: 16px;
-}
diff --git a/styles/index.css b/styles/index.css
deleted file mode 100644
index be6e5a9..0000000
--- a/styles/index.css
+++ /dev/null
@@ -1,473 +0,0 @@
-:root {
-  --paper: #ffffff;
-  --ink: #000000;
-  --text-on-dark: #ffffff;
-  --accent-a: #ffffff;
-  --accent-b: #7ed7ff;
-  --accent-c: #ffd2ef;
-}
-
-* {
-  box-sizing: border-box;
-}
-
-html,
-body {
-  width: 100%;
-  height: 100%;
-  min-height: 100%;
-  margin: 0;
-}
-
-body {
-  overflow: hidden;
-  min-height: 100svh;
-  background: var(--paper);
-  color: var(--text-on-dark);
-}
-
-.ghost-topbar {
-  position: fixed;
-  top: 0;
-  left: 0;
-  z-index: 20;
-  width: 100%;
-  display: flex;
-  align-items: center;
-  justify-content: space-between;
-  gap: 1rem;
-  padding: max(0.72rem, env(safe-area-inset-top)) clamp(0.85rem, 2.5vw, 2rem) 0.72rem;
-  color: #e8f7ff;
-  isolation: isolate;
-  pointer-events: auto;
-}
-
-.ghost-topbar::before {
-  content: "";
-  position: absolute;
-  inset: 0;
-  z-index: -2;
-  background:
-    linear-gradient(95deg, rgba(9, 14, 22, 0.84), rgba(7, 9, 14, 0.8)),
-    radial-gradient(circle at 14% 40%, rgba(74, 185, 242, 0.22), transparent 55%),
-    radial-gradient(circle at 78% 18%, rgba(255, 91, 191, 0.2), transparent 56%);
-  border-bottom: 1px solid rgba(132, 193, 216, 0.44);
-  backdrop-filter: blur(8px) saturate(130%);
-}
-
-.ghost-topbar::after {
-  content: "";
-  position: absolute;
-  inset: 0;
-  z-index: -1;
-  background-image:
-    repeating-linear-gradient(
-      0deg,
-      rgba(200, 240, 255, 0.05) 0,
-      rgba(200, 240, 255, 0.05) 1px,
-      transparent 1px,
-      transparent 3px
-    ),
-    repeating-linear-gradient(
-      90deg,
-      rgba(110, 170, 200, 0.06) 0,
-      rgba(110, 170, 200, 0.06) 1px,
-      transparent 1px,
-      transparent 4px
-    );
-  mix-blend-mode: screen;
-  opacity: 0.9;
-  animation: topbarNoiseDrift 2.1s steps(4, end) infinite;
-}
-
-.ghost-topbar__progress {
-  margin: 0;
-  min-width: 4ch;
-  font-family: 'JetBrains Mono', monospace;
-  font-size: clamp(0.86rem, 1.9vw, 1rem);
-  letter-spacing: 0.14em;
-  line-height: 1;
-  color: #8fe8ff;
-  text-shadow: 0 0 0.65rem rgba(84, 203, 255, 0.52);
-  font-size: 2.1rem;
-}
-
-.ghost-topbar.is-unlocked .ghost-topbar__progress {
-  color: transparent;
-  background-image: linear-gradient(
-    120deg,
-    var(--accent-a) 0%,
-    var(--accent-b) 38%,
-    var(--accent-c) 72%,
-    var(--accent-a) 100%
-  );
-  background-size: 220% 100%;
-  -webkit-background-clip: text;
-  background-clip: text;
-  -webkit-text-fill-color: transparent;
-  animation: wordmarkGradientShift 2850ms linear infinite;
-  text-shadow: none;
-}
-
-.ghost-topbar__nav {
-  margin-left: auto;
-}
-
-.ghost-topbar__links {
-  list-style: none;
-  margin: 0;
-  padding: 0;
-  display: flex;
-  align-items: center;
-  gap: clamp(0.4rem, 1.2vw, 0.75rem);
-}
-
-.ghost-topbar__links a {
-  display: inline-flex;
-  align-items: center;
-  min-height: 2rem;
-  padding: 0.38rem 0.72rem;
-  border-radius: 0.36rem;
-  font-family: 'Outfit', sans-serif;
-  font-size: clamp(0.72rem, 1.3vw, 0.84rem);
-  font-weight: 700;
-  letter-spacing: 0.03em;
-  text-transform: uppercase;
-  color: #e9f2f6;
-  text-decoration: none;
-  border: 1px solid rgba(182, 221, 240, 0.32);
-  background: rgba(18, 29, 44, 0.46);
-  transition: transform 180ms ease, border-color 180ms ease, background 180ms ease;
-}
-
-.ghost-topbar__links a:hover,
-.ghost-topbar__links a:focus-visible {
-  transform: translateY(-1px);
-  border-color: rgba(173, 230, 255, 0.78);
-  background: rgba(30, 56, 79, 0.7);
-}
-
-.ghost-topbar__links a:focus-visible {
-  outline: 2px solid rgba(191, 239, 255, 0.95);
-  outline-offset: 1px;
-}
-
-.ghost-topbar__toggle {
-  display: none;
-  width: 2.45rem;
-  height: 2.45rem;
-  border: 1px solid rgba(191, 236, 255, 0.48);
-  border-radius: 0.4rem;
-  background: rgba(16, 24, 37, 0.66);
-  padding: 0.34rem;
-  align-items: center;
-  justify-content: center;
-  cursor: pointer;
-}
-
-.ghost-topbar__toggle span {
-  display: block;
-  width: 1.2rem;
-  height: 0.12rem;
-  border-radius: 999px;
-  background: #d9f2ff;
-  transition: transform 200ms ease, opacity 200ms ease;
-}
-
-.ghost-topbar.is-open .ghost-topbar__toggle span:nth-child(1) {
-  transform: translateY(0.37rem) rotate(45deg);
-}
-
-.ghost-topbar.is-open .ghost-topbar__toggle span:nth-child(2) {
-  opacity: 0;
-}
-
-.ghost-topbar.is-open .ghost-topbar__toggle span:nth-child(3) {
-  transform: translateY(-0.37rem) rotate(-45deg);
-}
-
-@import url('https://fonts.googleapis.com/css2?family=League+Script&family=Outfit:weght@400;600;700;800;900&family=JetBrains+Mono:wght@400;700&display=swap');
-
-h1 {
-  margin: 0 0 4px;
-  font-family: 'League Script', cursive;
-  font-size: 60px;
-  font-weight: 400;
-  line-height: 1;
-}
-
-.ghost-hero {
-  position: relative;
-  width: 100%;
-  height: 100vh;
-  min-height: 100svh;
-  min-height: 100dvh;
-  overflow: hidden;
-  background: var(--paper);
-  cursor: crosshair;
-}
-
-.ghost-hero__image,
-.ghost-hero__veil,
-.ghost-hero__wordmark {
-  position: absolute;
-  inset: 0;
-}
-
-.ghost-hero__image {
-  width: 100%;
-  height: 100%;
-  object-fit: contain;
-  object-position: center center;
-  user-select: none;
-  -webkit-user-drag: none;
-}
-
-.ghost-hero__veil {
-  z-index: 2;
-  width: 100%;
-  height: 100%;
-  touch-action: none;
-  pointer-events: none;
-}
-
-.ghost-hero__wordmark {
-  z-index: 3;
-  display: grid;
-  place-items: center;
-  text-decoration: none;
-  color: inherit;
-  pointer-events: none;
-  cursor: crosshair;
-}
-
-.ghost-hero__wordmark::after {
-  content: "";
-  position: absolute;
-  width: clamp(6.5rem, 18vw, 13rem);
-  height: 0.13rem;
-  border-radius: 999px;
-  background: currentColor;
-  opacity: 0;
-  transform: translateY(clamp(2.4rem, 4.8vw, 4.4rem)) scaleX(0.65);
-  transition: opacity 220ms ease, transform 220ms ease;
-}
-
-.wordmark__layer {
-  position: absolute;
-  margin: 0;
-  line-height: 0.88;
-  letter-spacing: -0.055em;
-  font-size: clamp(3.2rem, 11vw, 8rem);
-  white-space: nowrap;
-  text-align: center;
-  user-select: none;
-  padding: 1em;
-  font-weight: 900;
-  color: var(--text-on-dark);
-  transition:
-    color 280ms ease,
-    transform 220ms ease,
-    filter 220ms ease;
-}
-
-.ghost-hero.is-unlocked .ghost-hero__wordmark {
-  pointer-events: auto;
-  cursor: pointer;
-}
-
-.ghost-hero.is-unlocked {
-  cursor: pointer;
-}
-
-.ghost-hero.is-unlocked .ghost-hero__wordmark::after {
-  opacity: 0.92;
-  transform: translateY(clamp(2.4rem, 4.8vw, 4.4rem)) scaleX(1);
-}
-
-.ghost-hero.is-unlocked .ghost-hero__wordmark:hover .wordmark__layer,
-.ghost-hero.is-unlocked .ghost-hero__wordmark:focus-visible .wordmark__layer {
-  transform: translateY(-0.05em);
-}
-
-.ghost-hero.is-unlocked .ghost-hero__wordmark:focus-visible {
-  outline: none;
-}
-
-.ghost-hero.is-unlocked .wordmark__layer {
-  color: var(--ink);
-}
-
-/* Lightweight CSS-only hover cue: moving gradient on text when unlocked. */
-.ghost-hero.is-unlocked .ghost-hero__wordmark:hover .wordmark__layer,
-.ghost-hero.is-unlocked .ghost-hero__wordmark:focus-visible .wordmark__layer {
-  color: transparent;
-  background-image: linear-gradient(
-    120deg,
-    var(--accent-a) 0%,
-    var(--accent-b) 38%,
-    var(--accent-c) 72%,
-    var(--accent-a) 100%
-  );
-  background-size: 220% 100%;
-  background-position: 0% 0%;
-  -webkit-background-clip: text;
-  background-clip: text;
-  -webkit-text-fill-color: transparent;
-  animation: wordmarkGradientShift 2850ms linear infinite;
-  filter: drop-shadow(0 0 0.4rem rgba(255, 255, 255, 0.35));
-}
-
-.ghost-hero.just-unlocked .wordmark__layer {
-  animation: unlockBlink 400ms linear 1;
-}
-
-.ghost-hero.just-unlocked .ghost-hero__wordmark::after {
-  animation: unlockUnderlineBlink 400ms linear 1;
-}
-
-@keyframes unlockBlink {
-  0% { opacity: 1; }
-  15% { opacity: 0.15; }
-  30% { opacity: 1; }
-  45% { opacity: 0.25; }
-  60% { opacity: 1; }
-  100% { opacity: 1; }
-}
-
-@keyframes unlockUnderlineBlink {
-  0% {
-    opacity: 0;
-    transform: translateY(clamp(2.4rem, 4.8vw, 4.4rem)) scaleX(0.4);
-  }
-  30% {
-    opacity: 1;
-    transform: translateY(clamp(2.4rem, 4.8vw, 4.4rem)) scaleX(1.06);
-  }
-  60% {
-    opacity: 0.35;
-    transform: translateY(clamp(2.4rem, 4.8vw, 4.4rem)) scaleX(0.88);
-  }
-  100% {
-    opacity: 0.92;
-    transform: translateY(clamp(2.4rem, 4.8vw, 4.4rem)) scaleX(1);
-  }
-}
-
-@keyframes wordmarkGradientShift {
-  from {
-    background-position: 0% 0%;
-  }
-  to {
-    background-position: 100% 0%;
-  }
-}
-
-@media (max-width: 900px), (pointer: coarse) {
-  .ghost-topbar {
-    gap: 0.7rem;
-    padding-right: max(0.85rem, env(safe-area-inset-right));
-  }
-
-  .ghost-topbar__toggle {
-    display: inline-flex;
-  }
-
-  .ghost-topbar__nav {
-    position: fixed;
-    top: calc(max(0.72rem, env(safe-area-inset-top)) + 2.72rem);
-    right: max(0.85rem, env(safe-area-inset-right));
-    width: min(84vw, 19.5rem);
-    pointer-events: none;
-  }
-
-  .ghost-topbar__links {
-    display: grid;
-    gap: 0.42rem;
-    padding: 0.52rem;
-    border: 1px solid rgba(165, 223, 247, 0.52);
-    border-radius: 0.56rem;
-    background: rgba(7, 12, 19, 0.9);
-    box-shadow: 0 0.7rem 1.5rem rgba(0, 0, 0, 0.35);
-    transform-origin: top right;
-    transform: scale(0.96);
-    opacity: 0;
-    transition: opacity 170ms ease, transform 170ms ease;
-  }
-
-  .ghost-topbar__links a {
-    width: 100%;
-    justify-content: flex-start;
-    font-size: 0.74rem;
-  }
-
-  .ghost-topbar.is-open .ghost-topbar__nav {
-    pointer-events: auto;
-  }
-
-  .ghost-topbar.is-open .ghost-topbar__links {
-    opacity: 1;
-    transform: scale(1);
-  }
-
-  .wordmark__layer {
-    width: 90vw;
-    font-size: clamp(3.4rem, 22vw, 8rem);
-    letter-spacing: -0.05em;
-    left: 50%;
-    top: 50%;
-    transform: translate(-50%, -50%);
-  }
-
-  .ghost-hero.is-unlocked .ghost-hero__wordmark:hover .wordmark__layer,
-  .ghost-hero.is-unlocked .ghost-hero__wordmark:focus-visible .wordmark__layer {
-    transform: translate(-50%, calc(-50% - 0.05em));
-  }
-
-  .ghost-hero__wordmark::after {
-    width: 80vw;
-    transform: translateY(calc(clamp(2.4rem, 4.8vw, 4.4rem) + 50vh - 0.07rem)) scaleX(0.65);
-  }
-
-  .ghost-hero.is-unlocked .ghost-hero__wordmark::after {
-    transform: translateY(calc(clamp(2.4rem, 4.8vw, 4.4rem) + 50vh - 0.07rem)) scaleX(1);
-  }
-}
-
-@media (min-width: 901px) and (pointer: fine) {
-  .ghost-topbar.is-open .ghost-topbar__nav {
-    pointer-events: auto;
-  }
-}
-
-@keyframes topbarNoiseDrift {
-  0% {
-    transform: translate(0, 0);
-    opacity: 0.84;
-  }
-  50% {
-    transform: translate(-1.2px, 0.8px);
-    opacity: 0.95;
-  }
-  100% {
-    transform: translate(0.9px, -0.7px);
-    opacity: 0.86;
-  }
-}
-
-@media (prefers-reduced-motion: reduce) {
-  .wordmark__layer {
-    transition: none;
-    animation: none;
-  }
-
-  .ghost-hero__wordmark::after {
-    transition: none;
-    animation: none;
-  }
-
-  .ghost-hero.is-unlocked .ghost-hero__wordmark:hover .wordmark__layer,
-  .ghost-hero.is-unlocked .ghost-hero__wordmark:focus-visible .wordmark__layer {
-    animation: none;
-  }
-}
diff --git a/styles/pages.css b/styles/pages.css
new file mode 100644
index 0000000..c8714fb
--- /dev/null
+++ b/styles/pages.css
@@ -0,0 +1,427 @@
+/* Editorial homepage, footer band revision.
+ *
+ * The camera band is now a standalone wide SVG motif, not a crop from the hero image.
+ * It is wider than many viewports so it can be bottom-aligned and cropped by CSS.
+ */
+
+:root {
+  --gm-bg: #fd5503;
+  --gm-bg-deep: #df5512;
+  --gm-ink: #050505;
+  --gm-text: #071a33;
+  --gm-muted: #253a54;
+  --gm-green: #063f32;
+  --gm-yellow: #f3c747;
+  --gm-cream: #ffe7a8;
+  --gm-border: rgba(5, 5, 5, 0.22);
+  --page-max: 1440px;
+  --camera-band-min-width: 2400px;
+  --camera-band-height: clamp(10rem, 23vw, 21rem);
+  --serif: "Iowan Old Style", "Palatino Linotype", "Book Antiqua", Palatino, Georgia, serif;
+  --sans: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif;
+}
+
+* {
+  box-sizing: border-box;
+}
+
+html,
+body {
+  margin: 0;
+  padding: 0;
+}
+
+body {
+  background:
+    radial-gradient(circle at top right, rgba(255, 220, 150, 0.08), transparent 34%),
+    linear-gradient(180deg, var(--gm-bg) 0%, var(--gm-bg) 68%, var(--gm-bg-deep) 100%);
+  color: var(--gm-text);
+  font-family: var(--sans);
+}
+
+a {
+  color: inherit;
+  text-decoration: none;
+}
+
+button,
+a {
+  -webkit-tap-highlight-color: transparent;
+}
+
+.cta-inline > i {
+  font-variant-caps: all-petite-caps;
+  font-size: 1.4rem;
+}
+
+.editorial-homepage {
+  min-height: 100svh;
+  position: relative;
+  overflow: clip;
+  isolation: isolate;
+}
+
+.editorial-homepage--fixed-camera-band .camera-band {
+  position: fixed;
+}
+
+.editorial-homepage__shell {
+  max-width: var(--page-max);
+  margin: 0 auto;
+  padding: 1.5rem 1.5rem calc(var(--camera-band-height) * 0.72 + 2rem);
+  position: relative;
+  z-index: 2;
+}
+
+.editorial-homepage__header {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 1.5rem;
+  border-bottom: 1px solid var(--gm-border);
+  padding-bottom: 1rem;
+}
+
+.wordmark {
+  display: inline-flex;
+  flex-direction: column;
+  gap: 0.15rem;
+}
+
+.wordmark__title {
+  font-family: var(--serif);
+  font-size: clamp(2.25rem, 2vw + 1.8rem, 3.4rem);
+  font-weight: 700;
+  line-height: 0.98;
+  letter-spacing: -0.04em;
+  color: var(--gm-ink);
+}
+
+.wordmark__subtitle {
+  font-size: 0.82rem;
+  text-transform: uppercase;
+  letter-spacing: 0.1em;
+  color: var(--gm-text);
+  opacity: 0.78;
+}
+
+.editorial-homepage__nav {
+  display: flex;
+  align-items: center;
+  gap: 1.4rem;
+  flex-wrap: wrap;
+  justify-content: flex-end;
+}
+
+.navlink {
+  font-size: 1.05rem;
+  font-weight: 650;
+}
+
+.navlink:hover,
+.navlink:focus-visible {
+  text-decoration: underline;
+  text-underline-offset: 0.2em;
+}
+
+.cta-inline {
+  display: inline-flex;
+  align-items: center;
+  gap: 0.55rem;
+  padding: 0.9rem 1.4rem;
+  border: 1.5px solid rgba(5, 5, 5, 0.65);
+  color: var(--gm-ink);
+  font-weight: 700;
+}
+
+.editorial-homepage__hero {
+  display: grid;
+  grid-template-columns: minmax(280px, 1.06fr) minmax(340px, 0.94fr);
+  align-items: center;
+  gap: clamp(2rem, 4vw, 4rem);
+  padding-top: clamp(2rem, 4vw, 4rem);
+}
+
+.hero-copy {
+  max-width: 42rem;
+}
+
+.hero-copy__title {
+  margin: 0;
+  font-family: var(--serif);
+  font-weight: 700;
+  color: var(--gm-ink);
+  font-size: clamp(4.4rem, 9vw, 8.8rem);
+  line-height: 0.9;
+  letter-spacing: -0.06em;
+}
+
+.hero-copy__lead {
+  margin: 1.45rem 0 0;
+  max-width: 32rem;
+  color: var(--gm-text);
+  font-size: clamp(1.1rem, 1.5vw, 1.7rem);
+  line-height: 1.18;
+  font-weight: 600;
+}
+
+/* Inline links inside narrative copy */
+.hero-copy__lead a,
+.hero-copy__small a {
+  color: var(--gm-ink);
+  font-weight: 760;
+  text-decoration: none;
+  padding-inline: 0.08em;
+  background:
+    linear-gradient(180deg,
+      transparent 56%,
+      rgba(243, 199, 71, 0.72) 56%,
+      rgba(243, 199, 71, 0.72) 90%,
+      transparent 90%);
+}
+
+.cta-inline:hover,
+.hero-copy__lead a:hover,
+.hero-copy__small a:hover,
+.cta-inline:focus-visible,
+.hero-copy__lead a:focus-visible,
+.hero-copy__small a:focus-visible {
+  background: var(--gm-cream);
+  color: var(--gm-ink);
+  outline: 2px solid rgba(5, 5, 5, 0.55);
+  outline-offset: 0.12em;
+}
+
+/* Metadata/action links */
+.hero-copy__meta {
+  align-items: flex-start;
+}
+
+.hero-copy__meta span {
+  border-top: 0;
+  padding-top: 0;
+}
+
+.hero-copy__meta a {
+  display: inline-flex;
+  align-items: center;
+  gap: 0.38em;
+  color: var(--gm-text);
+  font-weight: 760;
+  text-decoration: none;
+  padding: 0.42rem 0.62rem;
+  border: 1px solid rgba(7, 26, 51, 0.36);
+  background: rgba(255, 231, 168, 0.16);
+  box-shadow: 3px 3px 0 rgba(5, 5, 5, 0.15);
+}
+
+.hero-copy__meta a::after {
+  content: "↗";
+  font-size: 0.9em;
+}
+
+.hero-copy__meta a:hover,
+.hero-copy__meta a:focus-visible {
+  background: var(--gm-cream);
+  color: var(--gm-ink);
+  border-color: rgba(5, 5, 5, 0.72);
+  transform: translate(-1px, -1px);
+  box-shadow: 5px 5px 0 rgba(5, 5, 5, 0.22);
+  outline: none;
+}
+
+a:focus-visible,
+button:focus-visible {
+  outline: 3px solid var(--gm-cream);
+  outline-offset: 3px;
+}
+
+.hero-copy__small {
+  margin: 0.8rem 0 0;
+  max-width: 32rem;
+  color: var(--gm-muted);
+  font-size: 1rem;
+  line-height: 1.45;
+}
+
+.hero-copy__actions {
+  display: flex;
+  align-items: center;
+  gap: 1rem 1.35rem;
+  margin-top: 2rem;
+  flex-wrap: wrap;
+}
+
+.btn-primary {
+  display: inline-flex;
+  align-items: center;
+  gap: 0.85rem;
+  padding: 1rem 1.35rem;
+  background: var(--gm-text);
+  color: var(--gm-cream);
+  font-size: 1.15rem;
+  font-weight: 700;
+  min-width: 15rem;
+  justify-content: center;
+  box-shadow: 0 14px 32px rgba(7, 26, 51, 0.16);
+}
+
+.link-secondary {
+  display: inline-flex;
+  align-items: center;
+  gap: 0.5rem;
+  font-weight: 700;
+  font-size: 1.08rem;
+  border-bottom: 1.5px solid currentColor;
+  padding-bottom: 0.18rem;
+}
+
+.hero-copy__meta {
+  display: flex;
+  gap: 1rem;
+  flex-wrap: wrap;
+  margin-top: 1.6rem;
+  color: var(--gm-muted);
+  font-size: 0.88rem;
+}
+
+.hero-copy__meta span {
+  border-top: 1px solid rgba(7, 26, 51, 0.24);
+  padding-top: 0.55rem;
+}
+
+.hero-figure {
+  position: relative;
+  min-height: 0;
+  margin: 0;
+}
+
+.hero-figure__frame {
+  position: relative;
+  max-width: 38rem;
+  margin-left: auto;
+}
+
+.hero-figure__image {
+  width: 100%;
+  display: block;
+  object-fit: cover;
+}
+
+.camera-band {
+  position: absolute;
+  left: 50%;
+  bottom: 0;
+  z-index: 1;
+  width: max(100vw, var(--camera-band-min-width));
+  height: var(--camera-band-height);
+  transform: translateX(-50%);
+  background-image: url("../images/homepage-camera-band.svg");
+  background-repeat: no-repeat;
+  background-position: center bottom;
+  background-size: 100% 100%;
+  pointer-events: none;
+}
+
+.camera-band::before {
+  content: "";
+  position: absolute;
+  inset: 0;
+  background: linear-gradient(180deg, rgba(242, 106, 27, 0) 0%, rgba(242, 106, 27, 0.08) 36%, rgba(223, 85, 18, 0.08) 100%);
+}
+
+@media (min-width: 1700px) {
+  :root {
+    --camera-band-min-width: 3000px;
+    --camera-band-height: clamp(15rem, 20vw, 28rem);
+  }
+
+  .editorial-homepage__shell {
+    max-width: 1600px;
+  }
+}
+
+@media (max-width: 900px) {
+  :root {
+    --camera-band-min-width: 1700px;
+    --camera-band-height: clamp(8.5rem, 28vw, 13rem);
+  }
+
+  .editorial-homepage__shell {
+    padding-bottom: calc(var(--camera-band-height) * 0.86 + 1.5rem);
+  }
+
+  .editorial-homepage__header {
+    align-items: flex-start;
+    flex-direction: column;
+  }
+
+  .editorial-homepage__nav {
+    justify-content: flex-start;
+  }
+
+  .editorial-homepage__hero {
+    grid-template-columns: 1fr;
+    align-items: start;
+    gap: 1.8rem;
+  }
+
+  .hero-figure__frame {
+    margin: 0 auto;
+    max-width: 30rem;
+  }
+}
+
+@media (max-width: 620px) {
+  :root {
+    --camera-band-min-width: 1250px;
+    --camera-band-height: 7.8rem;
+  }
+
+  .editorial-homepage__shell {
+    padding: 1rem 1rem calc(var(--camera-band-height) + 1.5rem);
+  }
+
+  .editorial-homepage__header {
+    gap: 1rem;
+  }
+
+  .editorial-homepage__nav {
+    gap: 0.8rem 1rem;
+  }
+
+  .navlink {
+    font-size: 0.98rem;
+  }
+
+  .cta-inline {
+    width: 100%;
+    justify-content: center;
+  }
+
+  .hero-copy__title {
+    font-size: clamp(3.6rem, 20vw, 5rem);
+  }
+
+  .hero-copy__lead {
+    font-size: 1.02rem;
+    max-width: 100%;
+  }
+
+  .hero-copy__small {
+    font-size: 0.95rem;
+  }
+
+  .hero-copy__actions {
+    align-items: stretch;
+    flex-direction: column;
+  }
+
+  .btn-primary {
+    width: 100%;
+  }
+
+  .hero-figure__frame {
+    max-width: 23rem;
+  }
+}
\ No newline at end of file
diff --git a/styles/references-list.css b/styles/references-list.css
new file mode 100644
index 0000000..e79a339
--- /dev/null
+++ b/styles/references-list.css
@@ -0,0 +1,314 @@
+/* References list — additions on top of pages.css / content-pages.css.
+ * Uses the project's real custom properties (--gm-*) so it inherits the
+ * warm editorial palette automatically. No new colors are introduced;
+ * testability badges and tag flags reuse --gm-yellow / --gm-green /
+ * --gm-muted / --gm-ink.
+ */
+
+.reference-stats {
+  opacity: 0.75;
+}
+
+/* Screen-reader-only text (used for the star rating's numeric equivalent) */
+.sr-only {
+  position: absolute;
+  width: 1px;
+  height: 1px;
+  padding: 0;
+  margin: -1px;
+  overflow: hidden;
+  clip: rect(0, 0, 0, 0);
+  white-space: nowrap;
+  border: 0;
+}
+
+/* ---- Section grouping (peer-reviewed / art / advocacy / other) ---- */
+
+.reference-group {
+  display: flex;
+  align-items: baseline;
+  gap: 0.5rem;
+  margin: 2.5rem 0 1rem;
+  padding-top: 1.25rem;
+  border-top: 1px solid var(--gm-border);
+  font-family: var(--sans);
+  font-size: 0.85rem;
+  font-weight: 780;
+  text-transform: uppercase;
+  letter-spacing: 0.14em;
+  color: var(--gm-text);
+  opacity: 0.7;
+}
+
+.reference-group:first-of-type {
+  margin-top: 1.5rem;
+  border-top: none;
+  padding-top: 0;
+}
+
+.reference-group__count {
+  margin-left: 0.15rem; /* fallback for engines without flexbox `gap` support */
+  font-weight: 600;
+  letter-spacing: 0;
+  text-transform: none;
+  opacity: 0.8;
+}
+
+.reference-group__list {
+  display: grid;
+  gap: 1.5rem;
+}
+
+/* ---- Card shell ---- */
+
+.reference-entry {
+  padding: 1.1rem;
+  border: 1px solid rgba(7, 26, 51, 0.32);
+  background: rgba(255, 231, 168, 0.13);
+  box-shadow: 4px 4px 0 rgba(5, 5, 5, 0.13);
+}
+
+.reference-entry__meta {
+  display: flex;
+  flex-wrap: wrap;
+  align-items: center;
+  gap: 0.4rem;
+  margin-bottom: 0.75rem;
+}
+
+.reference-entry__meta--topics {
+  margin-top: 0.85rem;
+  margin-bottom: 0;
+}
+
+.reference-entry h3 {
+  margin: 0 0 0.3rem;
+  font-size: clamp(1.25rem, 1vw + 1rem, 1.7rem);
+}
+
+.reference-entry__authors {
+  margin: 0 0 0.6rem;
+  color: var(--gm-muted);
+  font-size: 0.92rem;
+}
+
+/* ---- Tags & badges ---- */
+
+.tag {
+  display: inline-flex;
+  align-items: center;
+  padding: 0.3rem 0.6rem;
+  border: 1px solid rgba(5, 5, 5, 0.24);
+  background: rgba(255, 231, 168, 0.4);
+  color: var(--gm-text);
+  font-family: var(--sans);
+  font-size: 0.72rem;
+  font-weight: 700;
+  letter-spacing: 0.03em;
+  line-height: 1;
+}
+
+.tag--topic {
+  text-transform: capitalize;
+  background: transparent;
+  color: var(--gm-muted);
+  font-weight: 600;
+}
+
+/* A tag no one else in the archive uses. */
+.tag--unique {
+  color: var(--gm-ink);
+  font-weight: 780;
+  border-color: rgba(5, 5, 5, 0.5);
+}
+
+/* The earliest entry using this tag — a small lineage marker, since we
+   deliberately don't offer click-to-filter on tags (see README). */
+.tag--new {
+  color: var(--gm-green);
+  font-weight: 780;
+  border-color: var(--gm-green);
+}
+
+.tag__flag {
+  display: inline-block;
+  margin-left: 0.4em;
+  padding: 0.05em 0.4em;
+  border-radius: 999px;
+  font-size: 0.85em;
+  font-weight: 780;
+  text-transform: uppercase;
+  letter-spacing: 0.03em;
+}
+
+.tag--unique .tag__flag {
+  background: var(--gm-yellow);
+  color: var(--gm-ink);
+}
+
+.tag--new .tag__flag {
+  background: transparent;
+  border: 1px solid var(--gm-green);
+  color: var(--gm-green);
+}
+
+.badge {
+  display: inline-flex;
+  align-items: center;
+  padding: 0.3rem 0.65rem;
+  border: 1px solid rgba(5, 5, 5, 0.4);
+  font-family: var(--sans);
+  font-size: 0.72rem;
+  font-weight: 780;
+  letter-spacing: 0.03em;
+  line-height: 1;
+  color: var(--gm-ink);
+}
+
+.badge--direct { background: var(--gm-yellow); }
+.badge--partial { background: transparent; border-color: var(--gm-green); color: var(--gm-green); }
+.badge--indirect,
+.badge--contextual { background: transparent; color: var(--gm-muted); }
+.badge--no { background: transparent; color: var(--gm-muted); text-decoration: line-through; }
+
+/* ---- Relevance stars (placeholder icon — see the <symbol> in the template) ---- */
+
+.reference-relevance {
+  display: inline-flex;
+  align-items: center;
+  gap: 0.1rem;
+  padding: 0.3rem 0.5rem 0.3rem 0.3rem;
+  border: 1px solid rgba(5, 5, 5, 0.24);
+  background: rgba(255, 231, 168, 0.4);
+}
+
+.star-icon {
+  fill: var(--gm-ink);
+  flex: none;
+}
+
+.star-icon--empty {
+  fill: none;
+  stroke: var(--gm-muted);
+  stroke-width: 1.5px;
+  opacity: 0.6;
+}
+
+/* ---- Full assessment (Demonstrated / relevance / testability / limits) ----
+ * A native <details> disclosure instead of four bordered boxes: collapsed by
+ * default so a page of 20+ entries stays scannable, and a flowing label:value
+ * list instead of a grid of little cards once it's open — reads the same
+ * whether the card is 320px or 1100px wide, no extra breakpoints needed. */
+
+.reference-entry__assessment summary {
+  display: inline-flex;
+  align-items: center;
+  gap: 0.4rem;
+  margin-top: 0.85rem;
+  padding: 0.4rem 0.75rem;
+  border: 1px solid rgba(5, 5, 5, 0.4);
+  background: rgba(255, 255, 255, 0.4);
+  font-family: var(--sans);
+  font-size: 0.78rem;
+  font-weight: 780;
+  text-transform: uppercase;
+  letter-spacing: 0.05em;
+  color: var(--gm-ink);
+  cursor: pointer;
+  list-style: none;
+}
+
+.reference-entry__assessment summary::-webkit-details-marker {
+  display: none;
+}
+
+.reference-entry__assessment summary::after {
+  content: "\25BE"; /* ▾ */
+  transition: transform 0.15s ease;
+}
+
+.reference-entry__assessment[open] summary {
+  margin-bottom: 0.9rem;
+}
+
+.reference-entry__assessment[open] summary::after {
+  transform: rotate(180deg);
+}
+
+.reference-entry__assessment summary:hover,
+.reference-entry__assessment summary:focus-visible {
+  background: var(--gm-cream);
+  border-color: rgba(5, 5, 5, 0.7);
+}
+
+.reference-entry__facts {
+  display: grid;
+  gap: 1rem;
+  margin: 0;
+  padding-top: 0.9rem;
+  border-top: 1px solid rgba(7, 26, 51, 0.18);
+}
+
+.reference-entry__fact {
+  display: grid;
+  gap: 0.2rem;
+}
+
+.reference-entry__fact dt {
+  margin: 0;
+  color: var(--gm-ink);
+  font-family: var(--sans);
+  font-size: 0.7rem;
+  font-weight: 780;
+  text-transform: uppercase;
+  letter-spacing: 0.08em;
+}
+
+.reference-entry__fact dd {
+  margin: 0;
+  font-size: 0.96rem;
+}
+
+.reference-entry__citation {
+  margin: 0.9rem 0 0.4rem;
+  color: var(--gm-muted);
+  font-size: 0.85rem;
+}
+
+.reference-entry__links {
+  margin: 0.5rem 0 0;
+  font-size: 0.92rem;
+}
+
+/* ---- Responsive tweaks ----
+ * Everything above already flows at any width (the label:value facts list,
+ * wrapping tag rows, and the global .content-page p max-width from
+ * content-pages.css keep prose readable even inside a very wide card) — the
+ * only real breakpoint needed is turning each fact's label+value from a
+ * stacked pair into a side-by-side spec-sheet row once there's room. */
+
+@media (min-width: 640px) {
+  .reference-entry__fact {
+    grid-template-columns: 11rem 1fr;
+    gap: 0.75rem 1rem;
+    align-items: baseline;
+  }
+}
+
+@media (min-width: 900px) {
+  .reference-entry {
+    padding: 1.5rem;
+  }
+}
+
+@media (max-width: 420px) {
+  .reference-entry {
+    padding: 0.9rem;
+  }
+
+  .tag,
+  .badge {
+    font-size: 0.68rem;
+    padding: 0.26rem 0.5rem;
+  }
+}
diff --git a/styles/report-page.css b/styles/report-page.css
new file mode 100644
index 0000000..74731c3
--- /dev/null
+++ b/styles/report-page.css
@@ -0,0 +1,164 @@
+/* report.html — additions on top of pages.css / content-pages.css.
+ * Reuses --gm-* tokens throughout; no new colors introduced beyond the
+ * existing palette. .resource-grid / .resource-card / .callout are already
+ * defined in content-pages.css and are reused as-is for the "Know more"
+ * links — this file only adds what's genuinely new: the audience cards,
+ * the receipt/secret-code callout, and the opsec checklist.
+ */
+
+/* ---- "Before you write anything" checklist ---- */
+
+.opsec-list {
+  max-width: 45rem;
+  margin: 1.25rem 0 0;
+  padding: 0;
+  list-style: none;
+  display: grid;
+  gap: 0.85rem;
+}
+
+.opsec-list li {
+  position: relative;
+  padding-left: 1.6rem;
+  color: var(--gm-text);
+  font-size: clamp(1.02rem, 0.25vw + 0.98rem, 1.18rem);
+  line-height: 1.5;
+}
+
+.opsec-list li::before {
+  content: "";
+  position: absolute;
+  left: 0;
+  top: 0.55em;
+  width: 0.6rem;
+  height: 0.6rem;
+  background: var(--gm-ink);
+  transform: rotate(45deg);
+}
+
+/* ---- Receipt / secret-code callout ---- */
+
+.receipt-callout {
+  display: flex;
+  flex-wrap: wrap;
+  align-items: center;
+  gap: 1.1rem;
+  max-width: 52rem;
+  margin-top: 1.4rem;
+  padding: 1.1rem 1.3rem;
+  border: 1.5px solid rgba(5, 5, 5, 0.55);
+  background: var(--gm-cream);
+}
+
+.receipt-callout__code {
+  flex: none;
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  padding: 0.6rem 1rem;
+  border: 1.5px dashed rgba(5, 5, 5, 0.55);
+  background: rgba(255, 255, 255, 0.55);
+  font-family: ui-monospace, SFMono-Regular, Menlo, Consolas, monospace;
+  font-size: 1.05rem;
+  font-weight: 700;
+  letter-spacing: 0.06em;
+  color: var(--gm-ink);
+}
+
+.receipt-callout p {
+  margin: 0;
+  max-width: none;
+  font-weight: 650;
+}
+
+/* ---- Audience cards ("Choose your questionnaire") ---- */
+
+.audience-grid {
+  display: grid;
+  grid-template-columns: repeat(2, minmax(0, 1fr));
+  gap: 1.1rem;
+  max-width: 64rem;
+  margin-top: 1.5rem;
+}
+
+.audience-card {
+  display: flex;
+  flex-direction: column;
+  gap: 0.7rem;
+  padding: 1.25rem;
+  border: 1px solid rgba(7, 26, 51, 0.32);
+  background: rgba(255, 231, 168, 0.13);
+  box-shadow: 4px 4px 0 rgba(5, 5, 5, 0.13);
+}
+
+.audience-card:hover,
+.audience-card:focus-visible {
+  background: var(--gm-cream);
+  border-color: rgba(5, 5, 5, 0.7);
+  transform: translate(-1px, -1px);
+  box-shadow: 6px 6px 0 rgba(5, 5, 5, 0.18);
+  outline: none;
+}
+
+.audience-card__icon {
+	display: inline-flex;
+	align-items: center;
+	justify-content: center;
+	width: 2.75rem;
+	height: 2.75rem;
+	/* border: 1px solid rgba(5, 5, 5, 0.4); */
+	background: var(--gm-cream);
+	color: var(--gm-ink);
+	border-radius: 10px;
+	margin-left: 1rem;
+}
+
+.audience-card__icon img {
+  width: 1.5rem;
+  height: 1.5rem;
+  display: block;
+}
+
+.audience-card h3 {
+  margin: 0;
+  font-family: var(--serif);
+  font-weight: 700;
+  font-size: clamp(1.15rem, 0.6vw + 1rem, 1.4rem);
+  line-height: 1.15;
+  letter-spacing: -0.02em;
+  color: var(--gm-ink);
+}
+
+.audience-card p {
+  margin: 0;
+  max-width: none;
+  color: var(--gm-muted);
+  font-size: 0.98rem;
+  line-height: 1.48;
+}
+
+.audience-card__cta {
+  margin-top: auto;
+  padding-top: 0.4rem;
+  font-family: var(--sans);
+  font-size: 0.8rem;
+  font-weight: 780;
+  text-transform: uppercase;
+  letter-spacing: 0.06em;
+  color: var(--gm-ink);
+}
+
+.audience-card__cta::after {
+  content: " \2197"; /* ↗ */
+}
+
+@media (max-width: 680px) {
+  .audience-grid {
+    grid-template-columns: 1fr;
+  }
+
+  .receipt-callout {
+    flex-direction: column;
+    align-items: flex-start;
+  }
+}
diff --git a/tests/e2e/face-matching.spec.js b/tests/e2e/face-matching.spec.js
index f3a3426..4460542 100644
--- a/tests/e2e/face-matching.spec.js
+++ b/tests/e2e/face-matching.spec.js
@@ -4,7 +4,7 @@ import { test, expect } from '@playwright/test';
  * End-to-End smoke test for analyze-panel entry point.
  */
 
-test.describe('Ghostati Analyze E2E', () => {
+test.describe('Ghostmaxxing Analyze E2E', () => {
   test('should open and close analyze panel', async ({ page }) => {
     // Capture console output from the page for debugging.
     page.on('console', msg => console.log(`BROWSER: ${msg.text()}`));
diff --git a/tests/e2e/overlay-mode.spec.js b/tests/e2e/overlay-mode.spec.js
index 54d7308..d69f252 100644
--- a/tests/e2e/overlay-mode.spec.js
+++ b/tests/e2e/overlay-mode.spec.js
@@ -8,7 +8,7 @@ function makeLandmarks478() {
   }));
 }
 
-test.describe('Ghostati Overlay Mode E2E', () => {
+test.describe('Ghostmaxxing Overlay Mode E2E', () => {
   test('cycles overlay modes, persists selection, renders mesh/bbox, and suppresses after save', async ({ page }) => {
     test.setTimeout(90000);
 
diff --git a/tests/unit/analyze-panel.test.js b/tests/unit/analyze-panel.test.js
index ed8414e..71140b2 100644
--- a/tests/unit/analyze-panel.test.js
+++ b/tests/unit/analyze-panel.test.js
@@ -57,7 +57,7 @@ describe('analyze-panel report', () => {
     });
 
     const text = generateReportText();
-    expect(text).toContain('### Ghostati - Analisi del trucco');
+    expect(text).toContain('### Ghostmaxxing - Analisi del trucco');
     expect(text).toContain('Volto rilevato: si');
     expect(text).toContain('Riconoscimento (face-api 2D)');
     expect(text).toContain('Embedder 3D (MediaPipe)');
diff --git a/tests/unit/engine.core.test.js b/tests/unit/engine.core.test.js
index 7174c73..189970d 100644
--- a/tests/unit/engine.core.test.js
+++ b/tests/unit/engine.core.test.js
@@ -132,7 +132,7 @@ describe('engine core exports', () => {
     window.Ghostati = {
       computeMatchState: vi.fn(() => 'matched')
     };
-    globalThis.Ghostati = window.Ghostati;
+    globalThis.Ghostmaxxing = window.Ghostati;
 
     const TinyFaceDetectorOptions = vi.fn(function TinyFaceDetectorOptions(opts) {
       Object.assign(this, opts);
diff --git a/tests/unit/engine.hasActivePlugin.test.js b/tests/unit/engine.hasActivePlugin.test.js
index 26068e1..7f95982 100644
--- a/tests/unit/engine.hasActivePlugin.test.js
+++ b/tests/unit/engine.hasActivePlugin.test.js
@@ -54,7 +54,7 @@ describe('engine.hasActivePlugin', () => {
     expect(hasActivePlugin()).toBe(true);
   });
 
-  it('returns false when Ghostati effect accessors are missing', () => {
+  it('returns false when Ghostmaxxing effect accessors are missing', () => {
     window.Ghostati = {};
     expect(hasActivePlugin()).toBe(false);
   });
diff --git a/tests/unit/ghostati.test.js b/tests/unit/ghostati.test.js
index f5c8510..703c090 100644
--- a/tests/unit/ghostati.test.js
+++ b/tests/unit/ghostati.test.js
@@ -2,14 +2,14 @@ import { describe, it, expect, beforeEach } from 'vitest';
 import '../../scripts/main.js';
 import { state } from '../../scripts/state.js';
 
-describe('Ghostati utils and state logic', () => {
-  let Ghostati;
+describe('Ghostmaxxing utils and state logic', () => {
+  let Ghostmaxxing;
 
   beforeEach(() => {
     window.localStorage.clear();
     state.db = { nextId: 0, faces: [] };
     state.MATCH_THRESHOLD = 0.58;
-    Ghostati = window.Ghostati;
+    Ghostmaxxing = window.Ghostati;
   });
 
   describe('Math utilities', () => {
@@ -17,32 +17,32 @@ describe('Ghostati utils and state logic', () => {
       const a = [0.1, 0.2, 0.3];
       const b = [0.1, 0.5, 0.3];
       // distance = sqrt((0.1-0.1)^2 + (0.2-0.5)^2 + (0.3-0.3)^2) = sqrt(0 + 0.09 + 0) = 0.3
-      const dist = Ghostati.distance(a, b);
+      const dist = Ghostmaxxing.distance(a, b);
       expect(dist).toBeCloseTo(0.3);
     });
 
     it('should return infinity for invalid descriptors', () => {
-      expect(Ghostati.distance(null, [0.1])).toBe(Number.POSITIVE_INFINITY);
-      expect(Ghostati.distance([0.1], [0.1, 0.2])).toBe(Number.POSITIVE_INFINITY);
+      expect(Ghostmaxxing.distance(null, [0.1])).toBe(Number.POSITIVE_INFINITY);
+      expect(Ghostmaxxing.distance([0.1], [0.1, 0.2])).toBe(Number.POSITIVE_INFINITY);
     });
 
     it('should correctly calculate avgPoint', () => {
       const points = [{ x: 10, y: 20 }, { x: 30, y: 40 }];
-      const avg = Ghostati.avgPoint(points);
+      const avg = Ghostmaxxing.avgPoint(points);
       expect(avg).toEqual({ x: 20, y: 30 });
     });
 
     it('should correctly calculate lerp', () => {
       const a = { x: 0, y: 0 };
       const b = { x: 100, y: 100 };
-      const result = Ghostati.lerp(a, b, 0.5);
+      const result = Ghostmaxxing.lerp(a, b, 0.5);
       expect(result).toEqual({ x: 50, y: 50 });
     });
 
     it('should correctly scale a point from center', () => {
       const center = { x: 50, y: 50 };
       const pt = { x: 60, y: 60 };
-      const scaled = Ghostati.scaleFrom(center, pt, 2);
+      const scaled = Ghostmaxxing.scaleFrom(center, pt, 2);
       // distance from center is 10,10. Scaled by 2 is 20,20. So 50+20=70
       expect(scaled).toEqual({ x: 70, y: 70 });
     });
@@ -50,15 +50,15 @@ describe('Ghostati utils and state logic', () => {
 
   describe('State and API accessors', () => {
     it('should expose threshold through the public API', () => {
-      expect(Ghostati.getMatchThreshold()).toBeCloseTo(0.58);
+      expect(Ghostmaxxing.getMatchThreshold()).toBeCloseTo(0.58);
     });
 
     it('should return a cloned DB snapshot', () => {
       state.db.faces = [{ id: 1, descriptor: [0.1, 0.2] }];
-      const copy = Ghostati.getDb();
+      const copy = Ghostmaxxing.getDb();
       expect(copy.faces.length).toBe(1);
       copy.faces.push({ id: 2, descriptor: [0.4, 0.5] });
-      expect(Ghostati.getDb().faces.length).toBe(1);
+      expect(Ghostmaxxing.getDb().faces.length).toBe(1);
     });
   });
 });
diff --git a/tests/unit/ghostati_additional.test.js b/tests/unit/ghostati_additional.test.js
index 5fe9438..466fa86 100644
--- a/tests/unit/ghostati_additional.test.js
+++ b/tests/unit/ghostati_additional.test.js
@@ -1,5 +1,5 @@
 /**
- * Additional unit tests for Ghostati utility functions.
+ * Additional unit tests for Ghostmaxxing utility functions.
  *
  * These tests aim to increase coverage beyond the existing core tests.
  * They focus on:
@@ -20,7 +20,7 @@ function getDbSnapshot() {
   return structuredClone(state.db);
 }
 
-const Ghostati = window.Ghostati;
+const Ghostmaxxing = window.Ghostati;
 
 /** Helper to create a mock CanvasRenderingContext2D with spy functions. */
 function createMockContext() {
@@ -42,7 +42,7 @@ function createMockContext() {
   return /** @type {CanvasRenderingContext2D} */ (ctx);
 }
 
-describe('Ghostati mathematical utilities', () => {
+describe('Ghostmaxxing mathematical utilities', () => {
   beforeEach(() => {
     state.db = { nextId: 0, faces: [] };
     state.MATCH_THRESHOLD = 0.58;
@@ -51,39 +51,39 @@ describe('Ghostati mathematical utilities', () => {
   it('distance returns Euclidean distance for equal-length vectors', () => {
     const a = [0, 0, 0];
     const b = [3, 4, 0];
-    expect(Ghostati.distance(a, b)).toBeCloseTo(5);
+    expect(Ghostmaxxing.distance(a, b)).toBeCloseTo(5);
   });
 
   it('distance returns POSITIVE_INFINITY for mismatched lengths', () => {
     const a = [1, 2];
     const b = [1];
-    expect(Ghostati.distance(a, b)).toBe(Number.POSITIVE_INFINITY);
+    expect(Ghostmaxxing.distance(a, b)).toBe(Number.POSITIVE_INFINITY);
   });
 
   it('avgPoint computes the correct centroid of point array', () => {
     const pts = [{ x: 0, y: 0 }, { x: 10, y: 20 }];
-    expect(Ghostati.avgPoint(pts)).toEqual({ x: 5, y: 10 });
+    expect(Ghostmaxxing.avgPoint(pts)).toEqual({ x: 5, y: 10 });
   });
 
   it('lerp interpolates correctly between two points', () => {
     const a = { x: 0, y: 0 };
     const b = { x: 10, y: 20 };
-    expect(Ghostati.lerp(a, b, 0.5)).toEqual({ x: 5, y: 10 });
+    expect(Ghostmaxxing.lerp(a, b, 0.5)).toEqual({ x: 5, y: 10 });
   });
 
   it('scaleFrom scales a point relative to a centre', () => {
     const centre = { x: 5, y: 5 };
     const pt = { x: 7, y: 9 };
     // vector from centre is (2,4); scaling by 2 => (4,8); result (9,13)
-    expect(Ghostati.scaleFrom(centre, pt, 2)).toEqual({ x: 9, y: 13 });
+    expect(Ghostmaxxing.scaleFrom(centre, pt, 2)).toEqual({ x: 9, y: 13 });
   });
 
   it('point utility returns an object with given coordinates', () => {
-    expect(Ghostati.point(3, 4)).toEqual({ x: 3, y: 4 });
+    expect(Ghostmaxxing.point(3, 4)).toEqual({ x: 3, y: 4 });
   });
 });
 
-describe('Ghostati drawing utilities with mocked canvas', () => {
+describe('Ghostmaxxing drawing utilities with mocked canvas', () => {
   let ctx;
   beforeEach(() => {
     ctx = createMockContext();
@@ -91,7 +91,7 @@ describe('Ghostati drawing utilities with mocked canvas', () => {
 
   it('drawClosedPath draws a filled and stroked polygon when styles provided', () => {
     const pts = [{ x: 0, y: 0 }, { x: 10, y: 0 }, { x: 10, y: 10 }];
-    Ghostati.drawClosedPath(ctx, pts, 'red', 'blue', 3);
+    Ghostmaxxing.drawClosedPath(ctx, pts, 'red', 'blue', 3);
     // Verify path construction calls.
     expect(ctx.beginPath).toHaveBeenCalled();
     expect(ctx.moveTo).toHaveBeenCalledWith(0, 0);
@@ -108,7 +108,7 @@ describe('Ghostati drawing utilities with mocked canvas', () => {
 
   it('drawOpenPath draws a dashed line when requested', () => {
     const pts = [{ x: 0, y: 0 }, { x: 5, y: 5 }, { x: 10, y: 0 }];
-    Ghostati.drawOpenPath(ctx, pts, 'green', 2, true);
+    Ghostmaxxing.drawOpenPath(ctx, pts, 'green', 2, true);
     expect(ctx.save).toHaveBeenCalled();
     expect(ctx.beginPath).toHaveBeenCalled();
     expect(ctx.setLineDash).toHaveBeenCalledWith([10, 8]);
@@ -119,7 +119,7 @@ describe('Ghostati drawing utilities with mocked canvas', () => {
   });
 });
 
-describe('Ghostati higher‑level utilities', () => {
+describe('Ghostmaxxing higher‑level utilities', () => {
   it('expandEyePolygon returns a polygon combining eyebrow and scaled eye', () => {
     const eye = [
       { x: 0, y: 0 },
@@ -131,7 +131,7 @@ describe('Ghostati higher‑level utilities', () => {
     ];
     const eyebrow = eye.map(p => ({ x: p.x, y: p.y - 5 }));
     const tone = { scale: 1.2, brow: 0.7, fill: 'rgba(0,0,0,0.2)', stroke: 'black' };
-    const poly = Ghostati.expandEyePolygon(eye, eyebrow, tone.scale, tone.brow);
+    const poly = Ghostmaxxing.expandEyePolygon(eye, eyebrow, tone.scale, tone.brow);
     expect(poly.length).toBeGreaterThanOrEqual(6);
     // Verify that the first three points are the transformed eyebrow points.
     expect(poly[0]).toMatchObject({ x: expect.any(Number), y: expect.any(Number) });