agentleFS
AbhillashjadhavAbhillashjadhav/AI-PM-essential-skills/CLAUDE.md

AI-PM-essential-skills

CLAUDE.md6 starsChanged 5 months agoMITView on GitHub

This header is a formatting convenience for the first prompt only, not a firing-frequency limit. Each skill keeps following its own SKILL.md trigger rules afterward, independently of the header and of each other: - model-complexity-router keeps firing its own compact line on every later distinct task handoff (not just the first). - pm-context-system keeps proposing its one-line "Log to memory" entry whenever a later decision or fact worth keeping surfaces (not just the first). - cli-over-mcp-auditor's proactive check is genuinely…

The file

We split this CLAUDE.md into the rules it gives an agent

2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every change to a SKILL.md goes through a PR.

    line 7
  2. Very specific:

    Every PR is reviewed using /pr-review before merge.

    line 8

≈ “Common” is an estimate: at least 1 in 5 of the 28,029 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

AI PM Skills Marketplace

What this repo is

The catalogue in .claude-plugin/marketplace.json contains 8 installable plugins for AI product managers. Three standalone utilities and the ContextPort toolkit also live here. Read README.md for product boundaries and docs/VALIDATION.md for the evidence available for each product; installation and live skill quality are not established by static checks.

PR rules (non-negotiable)

  • Every change to a SKILL.md goes through a PR. No direct pushes to main.
  • Every PR is reviewed using /pr-review before merge.
  • One concern per PR. Schema changes, logic changes, and doc changes are separate PRs.
  • PR description must state: what changed, why, and how to test the change.

Skill quality bar

A skill ships when:

  1. The SKILL.md has valid frontmatter (name, description, argument-hint)
  2. The skill body has at least 3 explicit hard rules
  3. A test input + expected output exists in the skill's README section
  4. The PR review returns APPROVE

What we do not ship

  • Skills that duplicate official Anthropic surfaces
  • Skills that produce unverifiable output (fabricated metrics, invented pricing)
  • Skills without explicit failure guardrails

Skill routing (fallback for not-yet-enabled sessions)

Plugin skills fire natively once .claude/settings.json (marketplace + enabledPlugins) is committed to the repo and the session trusts project settings. Sessions started before that commit, or where the plugin was installed mid-session rather than picked up from committed settings, won't have it in their invocable set yet. For those sessions, when a request matches one of the descriptions below, read that skill's SKILL.md from pm-tactical/skills/ and follow it exactly, as if it had fired natively:

  • Model choice / "which model should I use" / cost-per-task → pm-tactical/skills/model-complexity-router/SKILL.md
  • Generate an artifact and validate it against a spec / self-QA → pm-tactical/skills/builder-validator/SKILL.md
  • Improve or tune an existing prompt → pm-tactical/skills/prompt-optimizer-loop/SKILL.md
  • Audit MCP connectors / context filling up too fast → pm-tactical/skills/cli-over-mcp-auditor/SKILL.md
  • Set up project memory / remember stakeholders across sessions → pm-tactical/skills/pm-context-system/SKILL.md
  • Define what correct output means so a grader can be built / "build a grader" / "what should the eval check" → model-grader/skills/model-grader/SKILL.md

Session header protocol

On the first substantial prompt of a session, run whichever of model-complexity-router, cli-over-mcp-auditor, and pm-context-system have something to say at that moment and combine their lines into ONE compact header (max 4 lines) above the task response, instead of three separate interruptions. Each skill stays silent when it has nothing useful to say — an empty check contributes zero lines to the header, not a placeholder line.

This header is a formatting convenience for the first prompt only, not a firing-frequency limit. Each skill keeps following its own SKILL.md trigger rules afterward, independently of the header and of each other:

  • model-complexity-router keeps firing its own compact line on every later distinct task handoff (not just the first).
  • pm-context-system keeps proposing its one-line "Log to memory" entry whenever a later decision or fact worth keeping surfaces (not just the first).
  • cli-over-mcp-auditor's proactive check is genuinely once per session — its Step 0 doesn't re-fire after the first prompt.

Repo owner

Abhillash Jadhav — github.com/Abhillashjadhav

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.