AGENTS.md · diff
git:20260321.deb73d4 to git:20260629.ee6fb37
41 added, 26 removed. Audit A to A.
# AI Skills Library — Agent Instructions
## What This Repository Is
- This is the **universal AI skills library** — 204 production-ready skill packages across 13 professional domains with 559 Python automation tools, 17 AI agents, and 12 sample CI/CD workflows. It works with every major AI coding assistant: Claude Code, Cursor, Copilot, Codex, Windsurf, Cline, Aider, Goose, and more.
+ This is the **universal AI skills library** — **339 production-ready skill packages across 17 professional domains** with **787 Python automation tools**, 76 agents (incl. 8 cross-domain personas), 27 slash commands, and 8 CI/CD workflows. It works with every major AI coding assistant: Claude Code, Cursor, Copilot, Codex / ChatGPT, Gemini CLI, Windsurf, Cline, Aider, Goose, and more.
**This is NOT a traditional application.** It's a library of self-contained skill packages meant to be extracted and deployed by users into their own workflows.
## Supported Platforms
| Platform | Config File | Status |
|----------|------------|--------|
- | Claude Code | `CLAUDE.md` | Primary |
- | OpenAI Codex | `AGENTS.md` | Full |
- | Cursor | `.cursorrules` | Full |
+ | Claude Code | `CLAUDE.md` + `.claude-plugin/marketplace.json` | Primary |
+ | OpenAI Codex / ChatGPT | `AGENTS.md` (root + nested per-domain) | Full |
+ | Cursor | `.cursor/rules/*.mdc` (+ legacy `.cursorrules`) | Full |
+ | Gemini CLI | `GEMINI.md` + `.gemini/skills-index.json` | Full |
| GitHub Copilot | `.github/copilot-instructions.md` | Full |
| Windsurf | `.windsurfrules` | Full |
| Cline | `.clinerules` | Full |
| Goose | `.goosehints` | Full |
- | Aider | `AGENTS.md` | Full |
- | Jules | `AGENTS.md` | Full |
- | RooCode / Kilo Code | `AGENTS.md` | Full |
+ | Aider · Jules · RooCode / Kilo Code | `AGENTS.md` | Full |
- ## Repository Structure
+ > Codex and other AGENTS.md-aware agents also read **nested `AGENTS.md`/`CLAUDE.md` files** inside each
+ > domain folder for scoped guidance (e.g. `engineering/CLAUDE.md`, `project-management/CLAUDE.md`).
+ ## Repository Structure (domain → skill count)
+
```
- ├── engineering/ # 61 engineering skills + 177 Python tools
- ├── marketing/ # 35 marketing skills + 106 Python tools
- ├── product-team/ # 8 product skills + 15 Python tools
- ├── project-management/ # 25 PM skills + 53 Python tools
- ├── c-level-advisor/ # 26 C-level advisory skills + 73 Python tools
- ├── ra-qm-team/ # 21 regulatory/quality/compliance skills + 38 Python tools
- ├── business-growth/ # 17 business & growth skills + 48 Python tools
- ├── data-analytics/ # 5 data analytics skills + 16 Python tools
- ├── hr-operations/ # 4 HR operations skills + 12 Python tools
- ├── sales-success/ # 5 sales success skills + 15 Python tools
- ├── finance/ # 1 finance skill + 4 Python tools
- ├── agents/ # 11 production skill agents
- ├── scripts/ # Skill installer + utility scripts
- ├── standards/ # Best practices library
- └── templates/ # Reusable templates + 12 sample workflows
+ ├── engineering/ # 82 skills — architecture, fullstack, DevOps, security, AI/ML, data, agent tooling
+ ├── project-management/ # 66 skills — discovery, delivery, career, strategy-frameworks/, gtm/, Jira/Linear/Notion
+ ├── marketing/ # 39 skills — content, SEO/AEO, demand gen, brand, analytics
+ ├── c-level-advisor/ # 31 skills — CEO/CTO/CFO/CMO/CRO/CPO/COO/CHRO/CISO + CAIO/CDO/CCO/GC/VPE
+ ├── ra-qm-team/ # 27 skills — ISO 13485, MDR, FDA, SOC 2, GDPR, EU AI Act, NIS2, DORA, NIST CSF
+ ├── business-growth/ # 20 skills — CRO, pricing, revenue ops, commercial
+ ├── legal/ # 17 skills — EXPERIMENTAL: contracts, NDA, privacy, DPIA, breach response
+ ├── product-team/ # 13 skills — RICE, OKRs, UX research, product analytics, Apple HIG
+ ├── personal-productivity/ # 10 skills — resume, leads, meetings, invoices, email triage
+ ├── vertical-advisors/ # 7 skills — fintech, healthtech, edtech, ecommerce, proptech, climate-tech, marketplace
+ ├── data-analytics/ # 5 skills — analysis, BI, data science, ML ops
+ ├── sales-success/ # 5 skills — account exec, sales ops, solutions architect
+ ├── documents/ # 4 skills — docx, pdf, pptx, xlsx (stdlib OOXML)
+ ├── hr-operations/ # 4 skills — talent, people analytics, HRBP
+ ├── research/ # 4 skills — litreview, grants, patent, dossier
+ ├── finance/ # 3 skills — financial analysis, DCF, forecasting
+ ├── workflow/ # 2 skills — skill-router, handoff (meta-skills)
+ ├── agents/ # 68 cs-* domain agents + 8 cross-domain personas
+ ├── scripts/ # Manifest builder (build_manifest.py) + skill installer
+ ├── standards/ # Best-practices library
+ ├── templates/ # Reusable templates + sample workflows
+ └── cli/skills.json # Machine-readable manifest (regenerated by build_manifest.py)
```
## Skill Package Pattern
Every skill follows the same structure:
```
skill-name/
- ├── SKILL.md # Master documentation with workflows
+ ├── SKILL.md # Master documentation with workflows (progressive disclosure: lean Tier-1 + on-demand references)
├── scripts/ # Python CLI tools (standard library, no ML/LLM calls)
├── references/ # Expert knowledge bases
└── assets/ # User-facing templates
```
Knowledge flows: `references/` → `SKILL.md` workflows → `scripts/` execution → `assets/` templates.
## Quick Install
```bash
# Browse available skills
python scripts/skill-installer.py list
# Install a skill
python scripts/skill-installer.py install senior-fullstack --agent claude
python scripts/skill-installer.py install content-creator --agent cursor
- python scripts/skill-installer.py install ceo-advisor --agent vscode
+ python scripts/skill-installer.py install ceo-advisor --agent codex
# Keep skills up to date
python scripts/skill-installer.py update
```
+ After adding/removing/renaming any skill, regenerate the manifest so every platform surface stays in sync:
+
+ ```bash
+ python scripts/build_manifest.py # rewrites cli/skills.json + registry.json
+ ```
+
## Code Style
- **Python scripts:** Standard library only, CLI-first with argparse, JSON + human-readable output
- - **Markdown:** YAML frontmatter on all SKILL.md files, consistent formatting
+ - **Markdown:** YAML frontmatter on all SKILL.md files, descriptions ≤ ~60 tokens, consistent formatting
- **Commits:** Conventional commits — `feat(domain):`, `fix(tool):`, `docs(skill):`
## Key Principles
1. **Skills are products** — Each skill is deployable as a standalone package
2. **Documentation-driven** — Clear, actionable docs over generic advice
3. **Algorithm over AI** — Deterministic analysis (Python scripts) over LLM calls
4. **Template-heavy** — Ready-to-use templates users customize
5. **Self-contained** — No dependencies between skills
## Anti-Patterns
- Creating dependencies between skills
- Adding complex build systems or test frameworks
- Generic advice instead of specific, actionable frameworks
- LLM/ML calls in scripts (defeats portability)
- Over-documenting file structure