Product-Manager-Skills Arms Your Agent with 77 Battle-Tested PM Frameworks

Ask a generic agent to "write a PRD" and you get a generic document nobody trusts. Product-Manager-Skills is a 7k-star library from Dean Peters that puts 77 battle-tested PM frameworks into your agent — each skill carries not just the template but the failure modes and the judgment behind it, so the work improves and so do you.

Why This Skill Matters

The skills come in three types that build on each other: workflow skills run complete end-to-end processes like a full discovery cycle, interactive skills guide a decision — they ask 3-5 questions about your context, then offer numbered recommendations, each with a "use this when" rationale — and component skills are templates for specific deliverables like a user story. The README calls the coaching stance ABC, Always Be Coaching: if you want to learn the framework without doing the work, the skill teaches it either way.

The catalog reads like a PM career in order: problem-framing-canvas (MITRE's sequence), positioning-statement (Geoffrey Moore's template), prd-development, user-story (Mike Cohn format plus Gherkin acceptance criteria), press-release (Amazon Working Backwards), prioritization-advisor, a 14-skill market intelligence suite, and a lifecycle/EOL suite for the part of the job nobody trains — extending, replacing, or retiring a product.

Installation

For Claude Code, add the plugin marketplace:

claude /plugin marketplace add deanpeters/Product-Manager-Skills

Not on Claude Code? The README's get-started table covers the alternatives: Claude Desktop or Web users grab pm-skills-starter-pack.zip from GitHub Releases and upload the individual skill ZIPs to Claude Skills; Codex users take pm-skills-codex.zip, which installs .agents/skills and AGENTS.md. Themed packs exist too — Discovery, Strategy, Delivery, AI PM, Market Intel, and Lifecycle & EOL, plus 99-all-skills-pack.zip with all 77.

If you would rather try before wiring anything into an agent, the repo ships a Streamlit playground:

pip install -r app/requirements.txt
streamlit run app/main.py

Real Workflow: Pick the Right Prioritization Framework

1. Open your agent in a project and describe the decision instead of naming a framework:

We have 14 feature requests and one team. Help me figure out how to prioritize
them — ask me whatever you need to know first.

2. An interactive skill like prioritization-advisor asks 3-5 questions about your context, then recommends RICE, ICE, Kano, or the right alternative — with the rationale for why that framework fits here.
3. Execute the recommendation. Component skills like user-story produce the deliverable — Mike Cohn format with Gherkin acceptance criteria, anti-patterns included.
4. When you cannot tell which skill you need, intel-discipline-advisor triages the question on your desk into the right discipline, cadence, and executing skill.

Every skill follows the same SKILL.md structure — Frontmatter, Purpose, Input, Key Concepts, Application, Examples, Common Pitfalls, References — and each has a best_for field listing 3-5 scenarios, which makes scanning for fit fast.

Tips

• Skills deliberately avoid $ARGUMENTS templating: it only expands in Claude Code and renders as literal syntax everywhere else, so every skill has a plain-language Input section that works on any runtime.
• Arriving empty-handed is fine — the Input section uses the context you supply and walks you through the rest.
• The lifecycle suite right-sizes itself: every skill offers Light / Standard / Heavy settings, defaults to Standard, and names what a lighter choice leaves uncovered.
• For market research, the investigation chain — market-landscape-scan → competitive-research-snapshot → competitive-intel-watch → battle-card-builder — makes research a cadence where each skill consumes the previous one's schema.

When Not to Use This

The license is CC BY-NC-SA 4.0 — non-commercial use with share-alike, so repackaging the skills into a paid product or course is off the table without permission. These are PM thinking frameworks, not engineering tools: they will not write code or run analytics for you. And if your org already has house templates your stakeholders expect, adapting those may beat importing an outside framework.

See the leaderboard for more skills.