PM Skills
70 battle-tested product management frameworks, ready for Claude, Codex, ChatGPT, and any agent that can read structured knowledge.
- Source repo
- deanpeters/Product-Manager-Skills
- Stars
- ★ 7.5k
- Last updated
- 3d ago
- License
- NOASSERTION
- Primary language
- Shell
- FA score
- 53/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformChatGPT · Codex · Claude Code · Claude.ai
- Cost
- Free software; you pay for model usage
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A PM needs to quickly write a well-structured PRD; using the prd-development skill, they can produce a document with problem, personas, solution, metrics, and stories in 2-4 days.
- Not a fit if
- Anyone wanting to resell the skills in a paid product or course
- Non-technical users unwilling to set up Python and API keys
- Source review
- 53/100 · Major gaps 7 safety controls not found
What does this agent do, and when should you use it?
Product Manager Skills is a library of 70 battle-tested PM frameworks, templates, and interactive decision tools designed for both product managers and AI agents. It addresses the problem of generic AI output in product management by providing structured SKILL.md files that teach the 'why' behind each framework, failure modes to avoid, and judgment to apply them correctly. Skills are organized into three tiers: workflow skills (19) for end-to-end processes, interactive skills (27) with guided discovery via 3-5 questions, and component skills (24) for specific deliverables. The library covers problem framing, stakeholder alignment, customer discovery, prioritization, roadmap planning, PRD writing, market intelligence, AI product work, and career transitions. Installation options include a Claude Code plugin marketplace, ZIP packs for Claude Desktop/Web and Codex, and a Streamlit local playground. Licensed under CC BY-NC-SA 4.0, allowing commercial use in companies but prohibiting resale.
The repository provides a standardized set of skill files for AI agents (such as Claude Code, Codex, ChatGPT) and human PMs. Each skill file is a Markdown document with frontmatter metadata (name, description, type, intent, best_for, scenarios) and a fixed structure (Purpose, Input, Key Concepts, Application, Examples, Common Pitfalls, References). Workflow skills (e.g., discovery-process, roadmap-planning) guide end-to-end PM processes, interactive skills (e.g., prioritization-advisor) provide personalized recommendations through 3-5 questions, and component skills (e.g., user-story, prd-development) generate specific deliverables. It offers commands like claude /plugin marketplace add deanpeters/Product-Manager-Skills for Claude Code installation, and downloadable ZIP packs for Claude Desktop/Web and Codex. Users can browse, get recommendations, and run skills through the Streamlit app (streamlit run app/main.py), which supports multiple model providers (Anthropic, OpenAI, Ollama).
- A PM needs to quickly write a well-structured PRD; using the prd-development skill, they can produce a document with problem, personas, solution, metrics, and stories in 2-4 days.
- A team deciding feature priorities uses the prioritization-advisor skill; after answering 3-5 questions, it recommends RICE, ICE, Kano, or a suitable alternative.
- A new VP/CPO uses the executive-onboarding-playbook skill to follow a 30-60-90 day diagnostic plan for a smooth leadership transition.
- A product lead wants to conduct market competitive analysis; they use the market-intelligence-suite skills like competitive-analysis-process and battle-card-builder for evidence-based analysis.
- An AI product manager evaluates their team's approach using the ai-shaped-readiness-advisor skill to determine if they are automating tasks or redesigning workflows.
- A non-technical PM uses the Streamlit playground to describe their situation in plain English, get skill recommendations, and run skill directly.
How do you install or deploy this agent?
Installation varies by environment:
- Claude Code: Run
claude /plugin marketplace add deanpeters/Product-Manager-Skills. - Claude Desktop/Web: Download
pm-skills-starter-pack.zip(or themed packs), unzip, and upload individual skill ZIPs to Claude Skills. - Codex: Download
pm-skills-codex.zip, install to.agents/skillswithAGENTS.md. - Local playground: Clone repo, run
pip install -r app/requirements.txtandstreamlit run app/main.py, with API keys set as environment variables (Anthropic, OpenAI, or Ollama). For detailed guides, see docs/INSTALL-CLAUDE-DESKTOP.md, docs/INSTALL-CLAUDE-CODE.md, docs/INSTALL-CODEX.md. Note: Network access is required to download files from GitHub Releases.
How do you use this agent?
The basic workflow is: select a skill relevant to your scenario, read its SKILL.md, and follow the instructions. For interactive skills, answer the 3-5 questions and choose from numbered recommendations; the skill executes and explains the why. For workflow skills, follow the steps sequentially (e.g., run a full discovery cycle). In Claude Code, after installing the plugin, you can invoke skills via slash commands. In the Streamlit app, use 'Find My Skill' to describe your situation and get recommendations. Each skill has an Input section; you can provide full, partial, or no context, and the skill will guide you through the rest.
What are this agent's strengths and limitations?
- Comprehensive coverage of PM lifecycle from problem framing to market intelligence with 70 skills.
- Structured skill files are dual-purpose: they equip agents for professional work and teach human PMs the reasoning behind each framework.
- Multi-platform support: Claude Code, Claude Desktop/Web, Codex, and a Streamlit playground, with multiple installation paths.
- Actively maintained with frequent releases (v0.83 added a large Market Intelligence Suite).
- License is CC BY-NC-SA 4.0, which prohibits reselling the skills themselves, potentially limiting commercial redistribution.
- No empirical evidence of skill effectiveness or user adoption metrics in the repository.
- Reliance on external platforms (Claude, Codex) for runtime may cause compatibility issues if those platforms change.
- Streamlit playground requires API keys and Python setup, which may be a barrier for non-technical users.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| PM Skills This agent | 53 · Major gaps | Agent plugin / skillFree + model costs | ★ 7.5k | 3d ago | Shell | ChatGPT · Codex · Claude Code · Claude.ai |
| Lenny's Product Skills Pack | 38 · Major gaps | Agent plugin / skillFree | ★ 1.4k | 3mo ago | — | Claude Code |
| COMPASS Skills | 85 · Good | CLIFree | ★ 758 | today | Python | Codex · Claude Code |
| PM Skills — Professional Agent Skill Library | 74 · Some gaps | Agent plugin / skillFree + model costs | ★ 1.5k | today | HTML | ChatGPT · Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: The repository clearly states source attribution (e.g., MITRE ITK, Geoffrey Moore) and explains licensing and adaptation origins in the README. However, no information is provided on least privilege, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, or rollback mechanisms. Therefore, except for source attribution, all other criteria are scored 0 due to lack of evidence.
Evidence shows: The README and skill file structures are consistent, with uniform skill naming and format, indicating good self-consistency. However, dependency availability only mentions external frameworks (e.g., MITRE) without guarantees or verification. Failure messages are not clearly addressed in terms of error handling or user guidance. Thus, self-consistency is scored 2, while dependency availability and failure messages are scored 1.
Evidence shows: The README details multiple use cases and audiences (e.g., PMs, different platforms) and provides installation guides for various platforms. Capability boundaries are defined through skill types (workflow, interactive, component) and applicable scenarios. Trigger precision is reflected in skill descriptions and best_for fields. Environment fit covers multiple platforms (Claude, Codex, etc.). Therefore, audience and scenarios are scored 3, capability boundaries and trigger precision are scored 2, and environment fit is scored 3.
Evidence shows: Information architecture is clear with directories, categories, and documentation indexes. Installation notes are detailed with specific steps for different platforms. Naming is stable with consistent skill names. Examples and FAQ are abundant in README and docs. Known limitations only mention license restrictions, not other limitations. License is explicitly CC BY-NC-SA 4.0, but metadata is NOASSERTION, creating inconsistency. Versioning and changelog are detailed with version numbers and update records. Maintenance responsibility is clear via GitHub Issues and contribution guidelines. Thus, information architecture, install notes, examples and FAQ, versioning and changelog are scored 3; naming stability, license, and maintenance responsibility are scored 2; known limitations is scored 1.
Evidence shows: Output usability is high with structured outputs and examples. Marginal value is high with 70 skills covering broad PM scenarios. Cost-benefit is reasonable with free use and multiple installation methods, but no performance or resource consumption data is provided. Therefore, output usability and marginal value are scored 3, and cost-benefit is scored 2.
Evidence shows: Claim traceability is good with references to external frameworks and sources in the README. Cross-source corroboration is limited, only mentioning MITRE etc., without independent verification. Fact-inference separation is present in some skills (e.g., Fact/Inference/Assumption labels) but not consistently applied across all skills. Therefore, claim traceability is scored 2, cross-source corroboration is scored 1, and fact-inference separation is scored 2.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- License metadata is NOASSERTION, inconsistent with the CC BY-NC-SA 4.0 declared in README; actual license should be confirmed.
- No information on permission management, data flow transparency, or sensitive data handling; caution is advised when using.
- Dependencies on external frameworks (e.g., MITRE) are not verified, which may affect reliability.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.