Describe your manual process. Get a complete automation blueprint.
automation-blueprint is a Claude Skill that analyzes any manual business process and generates a structured automation blueprint with:
- Automation Score (0–100) across 5 dimensions
- Architecture diagram (Mermaid) with color-coded automation levels
- ROI calculation with conservative and optimistic scenarios
- Implementation plan phased by ROI priority
- Stack recommendation matched to your technical level
- Risk assessment and pre-requisite checklist
Built from patterns extracted from 21+ automations running in production.
Integrates three open-source methodologies:
| Methodology | Purpose | Link |
|---|---|---|
| ⚡ Stream Coding | 40/40/5/10/5 planning framework for implementation plans | github.com/frmoretto/stream-coding |
| 🔍 Clarity Gate | Epistemic verification — every number tagged [fact] / [estimate] / [hypothesis] |
github.com/frmoretto/clarity-gate |
| 🛡️ HardStop | Safety layer — human approval gates on destructive operations | github.com/frmoretto/hardstop |
# Clone the skill
git clone https://github.com/videomakingio-gif/claude-automation-blueprint.git
# Add to your Claude Code config
# In .claude/settings.json or project settings:
# "skills": ["path/to/claude-automation-blueprint"]For Cowork, download the .skill file from Releases and install it. The SKILL.md contains inline ROI formulas and cost tables — it works fully standalone.
Clone the repo into your project directory. The AI assistant will automatically discover the SKILL.md and reference files through its file system access.
Copy the content of SKILL.md as a system prompt or upload it as a file. For deeper analysis, also upload references/ and templates/. The skill is LLM-agnostic — all instructions, formulas, and structures work with any capable model.
Just describe your manual process in Italian or English:
"Every week I spend 6 hours preparing client reports. I pull data from Google Analytics, copy it into a spreadsheet, make charts, write commentary, and email the PDF."
or:
"Ogni settimana passo 6 ore a preparare report per i clienti. Prendo dati da Google Analytics, li copio in un Google Sheet, faccio grafici, scrivo un commento, e mando il PDF via email."
The skill will ask clarifying questions, then generate your complete blueprint.
See the examples/ directory for complete blueprints:
| Example | Score | Saved | ROI | Stack |
|---|---|---|---|---|
| LinkedIn Automation | 82/100 | 35+ hrs/month | 324% | Level 4 — Claude Code + Skills |
| Invoice Pipeline | 68/100 | 8 hrs/month | 302% | Level 2 — Claude Cowork + API |
| Content Pipeline | 71/100 | 20 hrs/month | 270% | Level 3 — Claude Code + CMS |
Structured questions to understand your process: frequency, duration, tools, costs, technical level. Max 5 questions at a time — no endless questionnaires.
- Automation Score (0–100) across 5 dimensions: repetitiveness, volume, data structure, API availability, economic value
- Friction mapping per step (🟢 low / 🟡 medium / 🔴 high)
Complete .md file with:
- Mermaid architecture diagram (color-coded by automation level)
- ROI calculation (conservative + optimistic, with epistemic tags)
- Phased implementation plan (following Stream Coding methodology)
- Stack recommendation (4 levels: no-code → full control)
- Safety layer with human approval gates (HardStop methodology)
- Risk assessment and pre-requisites
If automation doesn't make sense (score < 30), the skill tells you honestly and suggests manual optimizations instead. No upsell, no false promises.
| Feature | Generic ROI Calculator | This Skill |
|---|---|---|
| Process analysis | ❌ Input hours × cost | ✅ Step-by-step friction mapping |
| Architecture | ❌ None | ✅ Mermaid diagram with automation levels |
| ROI accuracy | ❌ Single number | ✅ Conservative + optimistic with ranges |
| Epistemic rigor | ❌ None | ✅ Every claim tagged [fact/estimate/hypothesis] |
| Safety | ❌ None | ✅ Human approval gates on destructive steps |
| Implementation plan | ❌ "Just automate it" | ✅ Phased plan with Stream Coding methodology |
| Honest "no" | ❌ Always says yes | ✅ Score < 30 = "don't automate" |
| Dimension | What it measures | 0–5 (low) | 16–20 (high) |
|---|---|---|---|
| Repetitiveness | How identical each execution is | Unique every time | Identical every time |
| Volume | How often the process runs | < 1/month | Daily+ |
| Data Structure | How structured the input/output is | Unstructured | Fully structured |
| API Availability | How accessible the tools are via API | No API | REST API, simple auth |
| Economic Value | Monthly savings potential | < €50/month | > €1000/month |
claude-automation-blueprint/
├── SKILL.md # Main skill file (self-contained)
├── references/
│ ├── automation-patterns.md # 11 patterns across 5 categories
│ ├── roi-framework.md # ROI formulas, cost tables, presentation rules
│ ├── stack-recommendations.md # 4 tech levels with architecture diagrams
│ ├── anti-patterns.md # 10 common mistakes to avoid
│ └── clarity-gate-integration.md # 9-checkpoint epistemic verification
├── templates/
│ ├── blueprint-template.md # Structured output template (IT/EN)
│ └── no-go-report-template.md # Template for Score < 30 (honest "no")
├── examples/
│ ├── example-linkedin-automation.md # Score 82 — LinkedIn management
│ ├── example-invoice-pipeline.md # Score 68 — Freelancer invoicing
│ └── example-content-pipeline.md # Score 71 — B2B content marketing
├── evals/
│ └── evals.json # 8 test cases with machine-readable assertions
├── assets/
│ └── hero.png # Repository hero image
├── .claude-plugin/
│ └── plugin.json # Marketplace metadata
├── CONTRIBUTING.md # Contribution guidelines
├── LICENSE # MIT
└── README.md # This file
EN: automation blueprint, process automation assessment, automation ROI calculator, Claude skill automation, business process automation, workflow automation analysis, should I automate, automation feasibility
IT: automazione processi aziendali, calcolo ROI automazione, processo manuale, conviene automatizzare, blueprint automazione, analisi processo, skill Claude automazione
By default, generated blueprints include an attribution footer with a link to the author's consulting page. You can disable the commercial link:
- Say "senza attribuzione" or "no attribution" during the conversation
- Or set
ATTRIBUTION: falseas context in your project settings
The attribution is optional and the skill works identically without it. The MIT license requires only the copyright notice in derivative works.
Most "automation assessments" are either generic ROI calculators or expensive consulting engagements. This skill bridges the gap: it produces a real architecture from a real process description, using patterns from production automation systems.
Built by Giovanni Liguori, AI Automation Architect with 21+ automations in production managing his entire digital presence via Claude.
Learn more:
- 📖 Blog post: Come funziona automation-blueprint
- 🏗️ Case Study: 21+ automazioni in produzione
- 📘 Claude Mastery Guide — How to get the most from Claude
- 🔗 5 Workflow Claude per risparmiare 40+ ore/mese
The blueprint is actionable on your own. But if you want someone to build it:
See CONTRIBUTING.md for guidelines. Issues, ideas, and PRs are welcome.
MIT — Giovanni Liguori, 2026