The open layer of how Tech Horizon Labs runs GEO / AI-visibility audits — a Claude Code skill suite, two original tools, and the method that ties them together.
The live measurement: the full measure asks all seven answer engines — ChatGPT · Claude · Gemini · Perplexity · Grok · DeepSeek · Google AI — whether they actually name you, and who they name instead. Up to 228 live answers per full measure, every one recorded verbatim, 3,500+ businesses on the public index. The free scan is the on-ramp: two buyer questions on two engines, no signup. This repo is the readiness diagnosis behind that score.
INSTALL · 17 skills into Claude Code, Codex, Cursor, Gemini CLI or any SKILL.md agent:
npx skills add techhorizonlabs/thl-open
A client-ready audit — composite score, evidence-backed findings, ranked fixes — produced by audit-report-kit in this repo. (Sample data is fictional.)
Search is splitting in two. A growing share of your future customers will never click a blue link — they'll get an answer straight from ChatGPT, Claude, Perplexity, or Google's AI Overview. GEO — Generative Engine Optimization — is how you stay inside that answer. This repo is how we measure it, engineer it, and prove the movement.
We're an Australian AI-visibility agency. This is the open layer of our practice: the method we actually use on client work, the tools we built ourselves, and our improvements to the open-source GEO suite we build on. The calibration data, client playbooks, and full engagement pack stay proprietary — but the method and the utilities are here, for free.
Same method, three depths — start wherever you are:
| ⚡ Measure visibility | areyoufoundbyai.com | Enter your URL → a live check of what AI answers when your buyers ask, in ~60 seconds. Free, hosted, nothing to install: two buyer questions on ChatGPT + Gemini, every answer kept verbatim. The trial runs the full seven-engine measure. The outcome. |
| 🛠 Diagnose readiness | this repo | Run the full six-dimension readiness audit in Claude Code — why you're visible or not, all the evidence, none of the black box. The inputs you control. |
| 🤝 Done for you | Tech Horizon Labs | We deploy it on your stack, with the calibration data and client playbooks that stay proprietary. |
Everything in this repo is organised around one opinionated workflow — the THL GEO Method. Three layers, each answering a different question:
flowchart TB
A["1 · Dimensional audit — geo-audit<br/>six GEO dimensions → a 0–100 composite + action plan"]
B["2 · External benchmark ★ — agent-readiness-scan<br/>Cloudflare's 0–100, tracked as a delta across re-audits"]
C["3 · Deliverable ★ — audit-report-kit<br/>a branded PDF + compile-checked JSON-LD to paste"]
A --> B --> C
A -.-> C
classDef thl fill:#6993ff,stroke:#1a1a2e,color:#ffffff;
class B,C thl;
The discipline that makes it repeatable: every score traces to evidence, the composite is a fixed formula, and you verify the artifacts — not that "a run happened." Full write-up in docs/THL-GEO-METHOD.md; the run-time QA lives in the audit checklist.
| Component | What it does |
|---|---|
skills/found-by-ai |
Wires any agent to the live measurement: run the free scan for a business, read the verdict and the rivals AI names instead, hand back the fix plan, and work the fix-and-re-measure loop over MCP for monitored sites. |
skills/agent-readiness-scan |
Turns Cloudflare's public isitagentready.com check into audit-grade artifacts — a fixed-schema CSV, the 0–100 score, and the evidence behind it. The independent benchmark the dimensional audit can't give itself. |
tools/audit-report-kit |
Turns an audit's JSON into a polished, branded client PDF (react-pdf) plus compile-checked JSON-LD (schema-dts) — the schema the audit recommends, ready to paste. (That's the report pictured above.) |
docs/THL-GEO-METHOD.md |
The orchestration + scoring discipline above. |
A comprehensive set of Claude Code skills for Generative Engine Optimization — full audits, citability scoring, AI-crawler analysis, llms.txt, schema markup, brand-mention scanning, platform-specific optimization, technical SEO, content E-E-A-T, and report generation.
Honest attribution: the
geo-*skills are an improved fork ofgeo-seo-claudeby Zubair Trabzada (MIT). We've de-rotted them, fixed portability, and wired them into our method and tools — but the foundation is his good open-source work, and we credit it. SeeNOTICE.mdandCHANGELOG.md. Improving and crediting open source is the standard we hold ourselves to.
As agent skills (Claude Code, Codex, Cursor, Gemini CLI, or any SKILL.md agent). GitHub is the registry; this repo is the package:
# the whole suite
npx skills add techhorizonlabs/thl-open
# or just the live-measurement skill
npx skills add techhorizonlabs/thl-open --skill found-by-ai
# see what's in the box first
npx skills add techhorizonlabs/thl-open --listAs a Claude Code plugin (installs the whole suite). Run these as two separate
commands — paste the first, press Enter, then paste the second (pasting both at once
makes Claude Code treat the blob as one broken /plugin call):
/plugin marketplace add https://github.com/techhorizonlabs/thl-open
/plugin install thl-open@thl-open
Use the full
https://URL — the shortowner/repoform makes some setups clone via SSH and fail without an SSH key configured.
Or copy just the skills you want:
cp -R skills/agent-readiness-scan ~/.claude/skills/
cp -R skills/geo-audit skills/geo-crawlers skills/geo-schema ~/.claude/skills/Then, in Claude Code:
run an agent-readiness scan on example.com
do a GEO audit of example.com
The report kit (Node):
cd tools/audit-report-kit && npm install && npm run report # → out/*.pdf + out/*.jsonld👉 New here? Start with examples/ — copy-paste prompts for every
skill and one worked audit (fictional broker) you can read end to end.
examples/— prompts + a worked end-to-end audit.- the THL GEO Method — the three layers and the scoring discipline.
- Sources — the primary research the method is grounded in, and what we don't claim.
- How THL Open compares — an honest look at the other GEO tools (and where they're ahead).
- Skill-Authoring Standard — how we author skills, and the bar for PRs.
evals/— how we keep the audit repeatable.- Live research from the measurement engine — who wins the moment buyers compare you (1,001 high-intent SERPs: brands lose 67% of their own "vs" searches) · JS-injected schema is invisible to AI crawlers · Microsoft's GEO model, measured.
- Free agent-era tools — AI crawler checker · DNS-AID checker (the emerging IETF agent-discovery record) · IndexNow submitter.
Heads up: skills can run code and fetch the open web. Read a skill before you install it, and treat fetched web content as data, not instructions. There's no telemetry — nothing here phones home (
SECURITY.md). Our authoring standard bakes this in; we ask the same of contributions.
Tech Horizon Labs is a Claude-native GTM + AI-visibility studio. We deploy this infrastructure for real clients — insurance brokers, professional-services firms, and scale-ups — on Cloudflare, into the workspace they already run. This repo is the readiness diagnosis we can give away — and areyoufoundbyai.com is the live visibility measurement, free for anyone to run in 60 seconds. Readiness here; visibility there. (How they fit together.)
If it's useful, a ⭐ helps. Issues and PRs welcome — see CONTRIBUTING.md.
MIT. The geo-* suite retains its original MIT notice (Zubair Trabzada) per skills/geo/LICENSE-geo-seo-claude; everything else is © Tech Horizon Labs. Use it, fork it, ship it.