PM-Skills

A curated library of 68 best-practice, plug-and-play product management skills covering the complete product lifecycle - plus templates, workflows, and 200+ real-world sample outputs that set the quality bar.

Report a Bug · Request a Feature · Ask a Question

About • Install • Skills • Sub-Agents • Workflows • Learning • Status • Contributing • Community

• Quick Start
• Recent Updates
• The Big Idea
  • Who Is This For
  • Key Features
  • Skill Lifecycle Tools
  • Founded On
• Installation and Setup
• How Agent Skills Work
• The Skill Library
  • Foundation Skills
  • Discover Phase Skills
  • Define Phase Skills
  • Develop Phase Skills
  • Deliver Phase Skills
  • Measure Phase Skills
  • Iterate Phase Skills
  • Tool Family: Foundation Sprint
  • Tool Family: Design Sprint
  • Standalone Tool Skill
  • Utility Skills
  • Sub-Agents
  • Workflows
• Library Examples
• Learning and Resources
• Project Status
• Contributing
• Community
• FAQ
• About the Author

Quick Start

After installing, you'll have all 68 skills available (invoke any by name, like /pm-skills:deliver-prd, /pm-skills:define-hypothesis, /pm-skills:deliver-user-stories) plus 10 /workflow-* orchestrator commands and the /chain ad-hoc runner, templates, sub-agents, and 200+ sample outputs.

Claude Code (recommended):

/plugin marketplace add product-on-purpose/agent-plugins
/plugin install pm-skills@product-on-purpose

Already installed via the old pm-skills-marketplace? It keeps working - no action needed. To move to the new home, see the v2.21.0 release notes.

Cross-agent (Cursor, Copilot, Cline, and others via the open skills CLI):

npx skills add product-on-purpose/pm-skills

Clone or download:

git clone https://github.com/product-on-purpose/pm-skills.git

More resources:

• Getting Started Guide - Detailed walkthrough for new users covering clone, sync helper, and first skill run. The path to take if any of the quick start steps above leave gaps.
• Setup by Platform - Step-by-step install for Claude.ai, Codex, Cursor, Windsurf, GitHub Copilot, VS Code extensions, and ChatGPT.
• Quickstart Reference - Short-form reference card for users who just need the commands.

(back to top)

Recent Updates

The companion pm-skills-mcp server is in the v2.9.x maintenance line (latest v2.9.3) and remains available on npm for clients that need MCP transport. The catalog is frozen at the v2.9.2 build; subsequent v2.9.x patches do not change the catalog. Active development on the MCP server is paused; security patches and critical bug fixes will continue.

Recommendation: For new users, we recommend using the plugin. See Installation and Setup for other options.

What's New

Full history: CHANGELOG.md or all GitHub Releases.

(back to top)

The Big Idea

Stop prompt-fumbling. Start shipping. Every time you ask an AI to help with product management, you start from zero. Generic responses. Inconsistent formats. Missing critical sections. Hours lost to repetitive prompt crafting.

PM-Skills gives your AI instant access to professional frameworks refined across hundreds of product launches, production-ready templates that capture institutional PM knowledge, and real-world examples that set the quality bar.

Who Is This For

Key Features

• ✅ 68 Production-Ready Skills covering the complete product lifecycle (30 phase skills + 11 foundation skills + 12 utility skills + 15 tool skills for structured workshop methodologies)
• ✅ Triple Diamond Framework organizing Discover, Define, Develop, Deliver, Measure, and Iterate phases
• ✅ Tool Families for the Foundation Sprint (2-day strategic alignment) and Design Sprint (5-day prototype-and-test) workshop methodologies
• ✅ 5 Active Orchestration Sub-Agents (pm-critic, pm-skill-auditor, pm-changelog-curator, pm-release-conductor, pm-workflow-orchestrator) for Claude Code, with dispatch skills extending the pattern to Codex, Cursor, Windsurf, Copilot, and Gemini CLI
• ✅ 12 Workflows for common PM processes including Feature Kickoff, Lean Startup, Triple Diamond, and foundation-to-design end-to-end arc
• ✅ 10 Workflow Commands for Claude Code users - plus every skill invocable directly by name
• ✅ Auto-Discovery via AGENTS.md in GitHub Copilot, Cursor, and Windsurf
• ✅ Agent Skills Spec compliant - works across AI assistants
• ✅ Apache 2.0 Licensed for commercial and personal use

Skill Lifecycle Tools = Custom Skill Builder

PM-Skills includes three utility skills that form a complete Create - Validate - Iterate lifecycle for building and managing skills:

%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#fffbeb', 'primaryBorderColor': '#fde68a', 'lineColor': '#92400e'}}}%%
flowchart LR
    classDef create fill:#ea580c,stroke:#c2410c,color:#fff,font-weight:bold
    classDef validate fill:#0284c7,stroke:#0369a1,color:#fff,font-weight:bold
    classDef decision fill:#374151,stroke:#1f2937,color:#fff
    classDef ship fill:#166534,stroke:#14532d,color:#fff,font-weight:bold
    classDef iterate fill:#7c3aed,stroke:#6d28d9,color:#fff,font-weight:bold

    Create["utility-pm-skill-builder<br/>Create new skill<br/>from an idea"]:::create
    Validate["utility-pm-skill-validate<br/>Audit structure<br/>and quality"]:::validate
    Decision{"Findings?"}:::decision
    Ship["Ship<br/>to library"]:::ship
    Iterate["utility-pm-skill-iterate<br/>Apply targeted<br/>improvements"]:::iterate

    Create --> Validate
    Validate --> Decision
    Decision -- "PASS" --> Ship
    Decision -- "WARN / FAIL" --> Iterate
    Iterate --> Validate

Why this matters: Skills are living artifacts that evolve. The builder creates them, the validator catches drift and quality gaps, and the iterator applies fixes. Together they keep the library consistent as it grows.

🔗 More resources:

• PM Skill Lifecycle Guide - Full workflow for creating, validating, and iterating skills, including the builder-validator-iterator cycle with worked examples.
• Creating PM Skills - Authoring guide for contributing new skills to the library.
• Skill Template - The canonical three-file template every skill must follow; copy this as your starting point.

Founded On

• Triple Diamond Framework - Six-phase product cycle (extends Design Council's Double Diamond): Discover, Define, Develop, Deliver, Measure, Iterate
• Opportunity Solution Trees (Teresa Torres) - Outcome-driven discovery
• Jobs to be Done Framework - Customer motivation framework
• Foundation Sprint (Knapp/Zeratsky) - 2-day structured workshop methodology for early-stage strategic alignment: basics, differentiation, approach options, magic lenses, founding hypothesis, brief
• Design Sprint (Knapp/Zeratsky/Kowitz) - 5-day structured workshop methodology for prototype-and-test cycles: map-and-target, sketch, decide-and-storyboard, prototype-plan, test-and-score
• Architecture Decision Records (Michael Nygard format) - Technical decision documentation

(back to top)

Installation and Setup

Tool Compatibility

Featured Install Paths

Claude Code - Plugin Marketplace (recommended)

/plugin marketplace add product-on-purpose/agent-plugins
/plugin install pm-skills@product-on-purpose

Already installed via the old pm-skills-marketplace? It keeps working - no action needed. To move to the new home, see the v2.21.0 release notes.

All 68 skills and their slash commands become available immediately. No clone required.

Cross-Agent via skills CLI (Cursor, Copilot, Cline, and others)

npx skills add product-on-purpose/pm-skills

The open skills CLI from Vercel Labs scans the skills/ directory and installs all 68 skills into your agent's default skills directory. Works with Claude Code, Cursor, GitHub Copilot, Cline, and any other agent that supports the skills ecosystem. Discoverable via skills.sh/product-on-purpose/pm-skills.

Telemetry is anonymous and opt-out via DISABLE_TELEMETRY=1 or DO_NOT_TRACK=1.

Git Clone (manual path, everything included)

git clone https://github.com/product-on-purpose/pm-skills.git
cd pm-skills

Gives you the full repo: skill files, slash commands, sample library, workflows, and documentation. Use this path when you want local access to sample outputs, plan to customize skills, or prefer not to depend on a CLI.

Optional: after cloning, run ./scripts/sync-claude.sh (macOS/Linux) or ./scripts/sync-claude.ps1 (Windows) to populate .claude/skills/ for agents that use that discovery path.

Additional Install Methods

For Claude.ai, MCP clients, OpenCode, Windsurf, and ChatGPT, see the full Platform Setup Guide for step-by-step instructions.

More resources:

• Getting Started Guide - Detailed walkthrough for new users covering all install methods, the sync helper script, and your first skill run end-to-end.
• Quickstart Reference - Short-form reference card with just the essential commands.

(back to top)

How Agent Skills Work

Each skill is a self-contained instruction set in three files:

%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#f8fafc', 'primaryBorderColor': '#cbd5e1', 'lineColor': '#64748b'}}}%%
flowchart LR
    classDef user fill:#1e293b,stroke:#0f172a,color:#f8fafc,font-weight:bold
    classDef agent fill:#4f46e5,stroke:#3730a3,color:#fff,font-weight:bold
    classDef skillFile fill:#7c3aed,stroke:#5b21b6,color:#fff
    classDef templateFile fill:#0284c7,stroke:#0369a1,color:#fff
    classDef exampleFile fill:#059669,stroke:#047857,color:#fff
    classDef output fill:#166534,stroke:#14532d,color:#fff,font-weight:bold

    User["You<br/>deliver-prd 'topic'"]:::user
    Agent["Your Agent"]:::agent
    SKILL["SKILL.md<br/>the method"]:::skillFile
    TEMPLATE["TEMPLATE.md<br/>the structure"]:::templateFile
    EXAMPLE["EXAMPLE.md<br/>the quality bar"]:::exampleFile
    Output["Complete artifact"]:::output

    User -- invokes --> Agent
    Agent -- reads --> SKILL
    Agent -- fills --> TEMPLATE
    Agent -- mirrors --> EXAMPLE
    SKILL --> Output
    TEMPLATE --> Output
    EXAMPLE --> Output

skills/deliver-prd/
  SKILL.md                  # the method the agent reads
  references/
    TEMPLATE.md             # the structure the output follows
    EXAMPLE.md              # the worked example that anchors quality

When you run /pm-skills:deliver-prd "topic", the agent loads the skill, mirrors the example, fills the template, and produces a complete PRD. No prompt engineering required.

🔗 More resources:

• PM Skill Anatomy - Deep dive into the three-file structure (SKILL.md, TEMPLATE.md, EXAMPLE.md) and what each file contributes to output quality. Essential reading before authoring your first skill.
• Skill Finder - Decision guide for choosing the right skill when you know the outcome you need but aren't sure which skill produces it.
• PM Skill Comparisons - Side-by-side comparison of overlapping skills. Prevents the common mistake of reaching for problem-statement when opportunity-tree is the right fit.
• Sub-Agents - Claude Code also supports spawnable specialist agents for tasks that benefit from a dedicated agent context: adversarial review, catalog auditing, changelog curation, and release orchestration.

(back to top)

The Skill Library

At a Glance

%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#f0f4ff', 'primaryBorderColor': '#c7d2fe', 'lineColor': '#6b7280', 'clusterBkg': '#f8fafc', 'clusterBorder': '#e2e8f0'}}}%%
flowchart TB
    classDef discover fill:#0891b2,stroke:#0e7490,color:#fff,font-weight:bold
    classDef define fill:#7c3aed,stroke:#6d28d9,color:#fff,font-weight:bold
    classDef develop fill:#d97706,stroke:#b45309,color:#fff,font-weight:bold
    classDef deliver fill:#16a34a,stroke:#15803d,color:#fff,font-weight:bold
    classDef measure fill:#dc2626,stroke:#b91c1c,color:#fff,font-weight:bold
    classDef iterate fill:#9333ea,stroke:#7e22ce,color:#fff,font-weight:bold
    classDef tool fill:#0284c7,stroke:#0369a1,color:#fff
    classDef foundation fill:#059669,stroke:#047857,color:#fff
    classDef utility fill:#ea580c,stroke:#c2410c,color:#fff

    subgraph PHASE["Triple Diamond Phase Skills (30)"]
        direction LR
        DI["Discover<br/>5 skills"]:::discover
        DE["Define<br/>5 skills"]:::define
        DEV["Develop<br/>4 skills"]:::develop
        DEL["Deliver<br/>6 skills"]:::deliver
        ME["Measure<br/>6 skills"]:::measure
        IT["Iterate<br/>4 skills"]:::iterate

        DI --> DE --> DEV --> DEL --> ME --> IT
        IT -."feedback loop".-> DI
    end

    subgraph WORKSHOP["Workshop Tool Families (15 skills)"]
        direction LR
        FS["Foundation Sprint<br/>7 skills<br/>2-day strategic alignment"]:::tool
        DS["Design Sprint<br/>7 skills<br/>5-day prototype-and-test"]:::tool
        NV["note-and-vote<br/>1 skill<br/>standalone mechanic"]:::tool
        FS --> DS
    end

    subgraph SUPPORT["Cross-Cutting Capabilities (23 skills)"]
        direction LR
        FOUND["Foundation<br/>11 skills<br/>persona, lean-canvas,<br/>OKRs, meetings"]:::foundation
        UTIL["Utility<br/>12 skills<br/>meta-tooling"]:::utility
    end

    PHASE -.uses.-> SUPPORT
    WORKSHOP -.uses.-> SUPPORT

Foundation Skills - Cross-cutting capability (11)

Phase Skills: Triple Diamond

Discover - Find and assess the right problem (5)

Define - Frame the problem (5)

Develop - Explore solutions (4)

Deliver - Ship it (6)

Measure - Validate with data (6)

Iterate - Learn and improve (4)

🔗 More resources:

• Using Skills Guide - Practical guide to invoking skills across different agents, passing context effectively, and getting consistent outputs session to session.
• Using Workflows Guide - How to run multi-skill chains, handle handoffs between skills, and adapt workflows to your team's process.
• Recipes - Curated patterns for common PM scenarios: solo feature launch, quarterly OKR cycle, research-to-delivery arc.
• Prompt Gallery - Real prompts that produce excellent skill outputs, annotated with what makes each one effective.

(back to top)

Tool Family: Foundation Sprint (7)

A Foundation Sprint is a short, structured workshop that clarifies the problem space, target users, success metrics, constraints, and business context before creative exploration begins. It ensures teams start with shared understanding and crisp definitions so downstream design and product decisions move faster and with fewer reversals.

Run the full methodology with the foundation-sprint workflow.

%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#eff6ff', 'primaryBorderColor': '#bfdbfe', 'lineColor': '#3b82f6'}}}%%
flowchart LR
    classDef sprint fill:#0284c7,stroke:#0369a1,color:#fff,font-weight:bold
    classDef gate fill:#1d4ed8,stroke:#1e40af,color:#fff,font-weight:bold

    R["1. Readiness<br/>Go / No-Go check"]:::gate
    B["2. Basics<br/>Customer + Problem + Competition"]:::sprint
    Di["3. Differentiation<br/>2x2 advantage map"]:::sprint
    A["4. Approach Options<br/>3-5 high-level paths"]:::sprint
    M["5. Magic Lenses<br/>Score approaches"]:::sprint
    H["6. Founding Hypothesis<br/>Synthesize direction"]:::sprint
    Br["7. Brief<br/>One-page output"]:::gate

    R --> B --> Di --> A --> M --> H --> Br

🔗 More resources:

• Foundation Sprint Concept Primer - A Foundation Sprint is a 2-day structured workshop methodology developed by Jake Knapp and John Zeratsky. This primer explains what it is, when to use it, the 7-step sequence, and how it compares to a Design Sprint.
• Using Foundation Sprint - Operational guide for running the 7-skill Foundation Sprint sequence with an agent, including facilitation notes and output expectations for each step.
• Foundation Sprint FAQ - Answers to common questions about the methodology: how long it takes, who should be in the room, what to do with the brief output, and when not to run it.
• Foundation Sprint Cheat Sheet - One-page quick reference for the day-arc, skill sequence, and key outputs.
• Foundation Sprint Case Studies - Three fictional scenarios showing how different teams used the Foundation Sprint methodology at different stages.
• Foundation Sprint Recovery - What to do when a sprint produces a confusing founding hypothesis, the team disagrees on direction, or you want to redo a step.

(back to top)

Tool Family: Design Sprint (7)

A Design Sprint is a fast, structured process that helps teams understand a problem, explore solutions, build a prototype, and test it with real users in five days. It reduces risk by forcing alignment and validation before any heavy engineering investment.

Run the full methodology with the design-sprint workflow. Chain it after a Foundation Sprint with foundation-to-design.

%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#fdf4ff', 'primaryBorderColor': '#e9d5ff', 'lineColor': '#9333ea'}}}%%
flowchart LR
    classDef sprint fill:#7c3aed,stroke:#6d28d9,color:#fff
    classDef gate fill:#581c87,stroke:#3b0764,color:#fff,font-weight:bold
    classDef day fill:#6d28d9,stroke:#5b21b6,color:#fff

    R["Readiness<br/>Go / No-Go"]:::gate
    Br["Brief<br/>Goal + Questions"]:::sprint
    M["Day 1<br/>Map and Target"]:::day
    S["Day 2<br/>Sketch"]:::day
    D["Day 3<br/>Decide and Storyboard"]:::day
    P["Day 4<br/>Prototype Plan"]:::day
    T["Day 5<br/>Test and Score"]:::gate

    R --> Br --> M --> S --> D --> P --> T

🔗 More resources:

• Design Sprint Concept Primer - A Design Sprint is a 5-day structured workshop methodology developed at Google Ventures by Jake Knapp, John Zeratsky, and Braden Kowitz. This primer covers the methodology's heritage, the 5-day arc as implemented in this skill family, and how it differs from a physical in-person workshop.
• Using Design Sprint - Operational guide for running the 7-skill Design Sprint sequence with an agent, including what "realistic-enough" means for the prototype step and how to recruit for test day.
• Design Sprint FAQ - Questions about prototype fidelity, user recruitment, scoring results, and what to do if no clear winner emerges after testing.
• Design Sprint Cheat Sheet - One-page quick reference for the 5-day arc, skill sequence, and key outputs.
• Design Sprint Case Studies - Fictional scenarios applying the Design Sprint to B2C, B2B, and internal-tools product challenges.
• Design Sprint Recovery - What to do when prototype testing produces inconclusive results, the test fails entirely, or the team needs to iterate on the storyboard.
• Workshop Sprints vs Agile Sprints - The disambiguation every Scrum team needs before running their first Design Sprint. Two completely different processes, one unfortunately shared word.
• Workshop Method Comparison - When to use Foundation Sprint vs Design Sprint vs other structured workshop approaches; includes a decision matrix.

(back to top)

Standalone Tool Skill

Utility Skills - Meta-tooling (12)

Sub-Agents

Sub-agents are specialist agents that Claude Code can spawn as autonomous sub-tasks within a larger session. Unlike skills - which are instruction files your agent reads inline - a sub-agent is a separate agent instance launched for a focused job. It runs against the skills catalog, produces its output, and returns the result to the orchestrating session. The orchestrator continues with that output as context.

%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#f8fafc', 'primaryBorderColor': '#cbd5e1', 'lineColor': '#64748b', 'clusterBkg': '#f1f5f9', 'clusterBorder': '#cbd5e1'}}}%%
flowchart LR
    classDef orchestrator fill:#1e293b,stroke:#0f172a,color:#f8fafc,font-weight:bold
    classDef subagent fill:#4f46e5,stroke:#3730a3,color:#fff,font-weight:bold
    classDef catalog fill:#7c3aed,stroke:#5b21b6,color:#fff
    classDef output fill:#166534,stroke:#14532d,color:#fff,font-weight:bold

    O["Your Claude Code session<br/>(orchestrator)"]:::orchestrator
    subgraph SA["Sub-Agent (spawned instance)"]
        direction LR
        Agent["Specialist agent"]:::subagent
        Cat["Skills catalog"]:::catalog
        Agent -- reads --> Cat
    end
    Out["Focused output<br/>returned to session"]:::output

    O -- "spawns" --> SA
    SA --> Out
    Out -- "informs" --> O

For other platforms (Codex, Cursor, Windsurf, Copilot, Gemini CLI), each sub-agent ships as a paired utility skill with dispatch instructions that replicate the same workflow in standard skill invocation mode.

🔗 More resources:

• Active Orchestration Guide - What sub-agents are, how Claude Code spawns them, the dispatch skill pattern for other platforms, and invocation examples for all five sub-agents.
• v2.16.0 Release Notes - The release that introduced active orchestration, including the rationale for the sub-agent model and how it connects to the broader skill ecosystem.

(back to top)

Workflows (skill chaining)

Workflows combine multiple skills into guided, end-to-end processes that mirror how experienced product managers actually work. Each workflow provides a sequence of skills with handoff guidance between steps so context flows naturally from discovery through delivery.

Ad-hoc chains and building your own. When no curated workflow fits, /chain runs any ordered sequence of skills against shared context (ephemeral, checkpointed by default; it routes to the pm-workflow-orchestrator engine). If a chain proves reusable, utility-pm-workflow-builder guides you from that chain (or a fresh idea) to a complete draft workflow packet, staged for review before anything lands in the repo.

(back to top)

Library Examples

Why Examples Matter More Than You Think

When you ask an AI to write a PRD, the output quality is anchored to its training data average. That average is mediocre. The examples in this library are the floor, not the ceiling. They are the quality signal your agent calibrates against before writing anything.

There are two layers of examples:

1. Skill-level EXAMPLE.md - every skill ships with a worked example sitting right next to the SKILL.md. This is what the agent reads at the moment it runs the skill.
2. Library samples (library/skill-output-samples/) - 200+ full outputs across three narrative threads. These are for you, not just the agent: read them before a skill run to calibrate expectations, or browse them to understand what a skill you haven't used actually produces.

Varied Prompt Maturity to Reflect Reality

The samples in this library intentionally vary in how polished the prompts are... Some use carefully structured, multi-turn context; others are more direct single invocations. This was a deliberate choice for two reasons:

• To reflect real life. The AI tooling landscape is evolving fast. Product managers are figuring out what works in real time. A library full of "perfect" prompts would misrepresent how most people actually work with AI today.
• To lower the bar for experimentation. The goal of this library is to encourage you to try, not to intimidate you with a gallery of polished outputs. If a sample was produced with a rough prompt and still looks good, that's a feature - it means the skill is doing the heavy lifting.

You'll find a natural range across the samples:

• Direct invocations with minimal context (most Brainshelf samples, early-stage framing)
• Single-turn invocations with a short briefing paragraph (most Storevine samples)
• Multi-step structured invocations with full context blocks (most Workbench ADR and stakeholder samples)

Browse the library to find the level of maturity that matches your current practice - then experiment upward from there.

Three Narrative Threads Across Three Fictional Companies

The library isn't a random collection of samples. It's organized around three fictional companies that each operate at a different stage, in a different industry, with different PM challenges. Every sample belongs to one of these threads so you can follow a company's product story across multiple skills.

Brainshelf - Early-Stage B2C Founder

• The company: Brainshelf builds a personal knowledge product - a tool for people who capture notes, articles, and ideas across too many apps and an never find them later. Think early-stage B2C, one PM who is also the founder, zero product-market fit certainty.
• Why this thread exists: Early-stage PM work looks different. Hypotheses are looser. Personas aren't validated by a research team. Foundation Sprints and lean canvases matter more than full PRDs. The Brainshelf samples show how to use PM skills when you have more questions than answers.
• Best samples to start with: define-hypothesis, tool-foundation-sprint-basics, discover-interview-synthesis, tool-foundation-sprint-founding-hypothesis

Storevine - Mid-Stage E-Commerce PM

• The company: Storevine is a mid-market e-commerce SaaS platform. The PM running the checkout-conversion program is three years into the role, works with an engineering team of six, and is accountable to experiment velocity and revenue-per-visit metrics.
• Why this thread exists: Most PM skill work happens in the middle: running experiments, measuring results, writing PRDs for known problems, aligning stakeholders on tradeoffs. The Storevine samples represent the bread-and-butter PM output at a company with enough scale to run rigorous tests but without so much bureaucracy that every artifact becomes a political battle.
• Best samples to start with: deliver-prd, measure-experiment-design, measure-okr-grader, define-opportunity-tree

Workbench - Internal-Tools PM at a Growing Org

• The company: Workbench is the internal tooling and platform team at a 200-person organization moving from startup to scale. The PM's "users" are internal employees and the "product" is the internal tooling layer - blueprints, automation, integrations.