Add CLAUDE.md and AGENTS.md agent guidance

Introduce repo-level guidance for AI agents:
- CLAUDE.md: single source of truth (structure, design rules, versioning,
  operational procedures). 65 skills, 36 commands, 8 plugins.
- AGENTS.md: thin pointer to CLAUDE.md for non-Claude agents.
- .gitignore: stop ignoring CLAUDE.md (now public); ignore CLAUDE.local.md
  and keep ignoring _Internal/ (maintainer-only, never published).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Pawel Huryn
2026-06-04 12:57:59 +02:00
parent f9eaa51000
commit 2b4e4dc151
3 changed files with 115 additions and 1 deletions
+2 -1
View File
@@ -1,2 +1,3 @@
# Private maintainer-only files — never commit
_Internal/
CLAUDE.md
CLAUDE.local.md
+5
View File
@@ -0,0 +1,5 @@
# AGENTS.md
All agent guidance for this repository lives in **[CLAUDE.md](CLAUDE.md)** — that is the single source of truth.
Please read [CLAUDE.md](CLAUDE.md) before making changes. This file exists only so non-Claude agents that look for `AGENTS.md` are pointed to the same instructions; do not duplicate guidance here.
+108
View File
@@ -0,0 +1,108 @@
# CLAUDE.md
Guidance for AI agents (Claude Code, Cowork, and others) working in this repository. This file is the single source of truth for how the project is structured and maintained.
## Project Overview
**PM Skills** (`phuryn/pm-skills`) — a marketplace of **8 independent plugins** (65 skills, 36 commands) that bring structured product-management workflows to AI coding assistants. Built for Claude Code and Claude Cowork; the skills are also compatible with other agents (Gemini CLI, Cursor, Codex CLI).
Owner: Paweł Huryn — pawel@productcompass.pm — https://www.productcompass.pm
## Repo Structure
```
pm-skills/ <- repo root
├── .claude-plugin/marketplace.json <- root marketplace manifest (lists all 8 plugins)
├── .docs/images/ <- images used by README (webp, gif)
├── .gitattributes
├── .gitignore
├── CLAUDE.md <- this file (agent guidance, single source of truth)
├── AGENTS.md <- pointer to CLAUDE.md (for non-Claude agents)
├── CONTRIBUTING.md <- contributor guidelines
├── README.md <- public documentation (GitHub)
├── LICENSE <- MIT
├── validate_plugins.py <- plugin validator
└── pm-{name}/ <- 8 plugin directories
├── .claude-plugin/plugin.json <- per-plugin manifest
├── skills/{skill}/SKILL.md <- one folder per skill
├── commands/{command}.md <- one file per command
└── README.md <- per-plugin documentation
```
### The 8 plugins
| Plugin | Focus |
|--------|-------|
| `pm-product-discovery` | Ideation, experiments, assumption testing, prioritization, interview synthesis |
| `pm-product-strategy` | Vision, strategy/lean/business-model canvas, SWOT, PESTLE, Ansoff, Porter, monetization |
| `pm-execution` | PRDs, OKRs, roadmaps, sprints, pre-mortems, stakeholder maps, user stories |
| `pm-market-research` | Personas, segmentation, sentiment analysis, competitive analysis, market sizing |
| `pm-data-analytics` | SQL query generation, cohort/retention analysis |
| `pm-go-to-market` | GTM strategy, growth loops, motions, beachhead segments, ICPs |
| `pm-marketing-growth` | Marketing ideas, value-prop statements, North Star metrics, naming, positioning |
| `pm-toolkit` | Resume review, NDA drafting, privacy policy, grammar/flow checking |
## Key Design Rules
- **Skills = nouns/concepts.** Frameworks and analytical knowledge Claude auto-loads when the topic matches (`lean-canvas`, `pre-mortem`, `market-sizing`).
- **Commands = verbs.** User-triggered workflows that chain one or more skills (`/write-prd`, `/discover`, `/plan-launch`).
- **No cross-plugin references.** Commands suggest follow-ups in natural language only ("Want me to design growth loops?"). Never hard-reference a command from another plugin — plugins install independently, so a hard reference can break.
- **Intra-plugin "Uses" references are fine** — skills and commands in the same plugin always ship together.
- Commands use a single `$ARGUMENTS` placeholder. Skills need no placeholders (they read context from the conversation).
- **Frontmatter required:** Skills need `name` + `description`; commands need `description` + `argument-hint`.
- A skill's `name` **must match its directory name**.
- Skills can be force-loaded with `/plugin-name:skill-name` or `/skill-name`.
- Keep frontmatter lean (always loaded); put detail in the SKILL.md body (loaded when triggered) — progressive disclosure.
## What's Visible Where
| Location | Visible in | Notes |
|----------|-----------|-------|
| `marketplace.json``description` | Cowork marketplace browser, Claude Code | One-liner for the whole marketplace |
| `plugin.json``description` | Cowork plugin list, Claude Code | Per-plugin summary; concise and functional |
| `SKILL.md` frontmatter → `description` | Cowork skill list, Claude auto-loading | Include trigger phrases so Claude loads the skill at the right time |
| Command frontmatter → `description` + `argument-hint` | Cowork and Claude Code (typing `/`) | Short and actionable |
| `README.md` (repo root) | GitHub only | Full docs; not loaded by Claude at runtime |
Descriptions in `plugin.json` and the repo `README.md` should stay aligned (identical text).
## Versioning
- All versions are currently **1.0.1**`marketplace.json` and all 8 `plugin.json` files.
- **Keep every version in sync.** There is no independent per-plugin versioning.
- Bump any `plugin.json` → also bump `marketplace.json`, and vice-versa (bump all 8 to match).
## Article Links in Skills (Further Reading)
- Mapped skills end with a `### Further Reading` section linking to relevant Product Compass articles.
- **Tone must stay neutral** — no promotional language, no CTAs, no "subscribe"/"check out". Just the article title and URL.
- Claude surfaces these links based on conversational relevance, not on every response.
- Posts whose title contains "Masterclass" or "Course" are video courses — tag them `(video course)`.
## Operational Procedures
### After any skill/command change
1. Run `python3 validate_plugins.py` from the repo root to check all plugins.
2. If skills/commands were added or removed, update the counts in `README.md`.
3. If totals changed, update the count in the `marketplace.json` description.
4. Bump versions across all manifests (see Versioning).
### After a description change
- A `plugin.json` description changed → check whether `README.md` needs the same edit (they stay aligned).
- A `SKILL.md` description changed → no other sync needed (it's the single source for that skill).
## Validation
`validate_plugins.py` checks: `plugin.json` required fields / name match / semver / author / keywords; skill frontmatter and name-matches-directory; command frontmatter (`description` + `argument-hint`); README presence; and intra-plugin command→skill references.
```
python3 validate_plugins.py
```
## What to Suggest After Completing Work
Offer relevant follow-ups:
- After structural changes: "Want me to run the validator?"
- After adding/removing skills or commands: "Should I update the counts in README.md and marketplace.json?"
- After editing descriptions: "Should I sync this to README.md / plugin.json?"
- After any repo change: "Want me to bump the version?"