Skip to content

Quickstart

  • 68 shipped PM skills in skills/ (30 phase skills across 6 phases, 11 foundation skills, 12 utility skills, 15 tool skills)
  • 11 slash-command docs in commands/ (10 /workflow-* orchestrator commands + the /chain command)
  • 12 Workflows for multi-skill processes (Triple Diamond, Lean Startup, Feature Kickoff, and 9 more)

Install from the plugin marketplace, no clone required:

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

All 68 skills are available immediately. Invoke any by name, for example /pm-skills:deliver-prd. Already on the old pm-skills-marketplace? It keeps working; see Setup by Platform to switch homes.

Cross-agent (Cursor, GitHub Copilot, Cline, and others)

Section titled “Cross-agent (Cursor, GitHub Copilot, Cline, and others)”
Terminal window
npx skills add product-on-purpose/pm-skills

Installs all 68 skills into your agent’s default skills directory via the open skills CLI. No clone, no sync.

The CLI’s --skill flag installs individual skills instead of the whole library, which is supported. Note that skills route to each other: a build-risk review ends by sending you to a hypothesis skill, a survey analysis points at experiment design. Skills that route as part of their output detect a missing destination and inline the minimum version of it rather than leaving a dead pointer, but the handoffs are seamless only with the full library. Installing a subset works best when you take the skills a routing table sends you to as well.

  1. Go to Settings > Capabilities (Desktop) or Project Settings > Add Files (Claude.ai)
  2. Upload the latest release ZIP (pm-skills-vX.X.X.zip) from Releases
  3. Skills are now available in your conversations
Terminal window
git clone https://github.com/product-on-purpose/pm-skills.git

Or download and extract the latest ZIP from Releases. Point any other agent to AGENTS.md for skill discovery; each skill is self-contained in skills/{skill-name}/SKILL.md (for example skills/deliver-prd/SKILL.md).

More detail: see the full getting-started guide for the long-form walkthrough.

Confirm the install landed before you start building:

  • Claude Code: run /plugin list to confirm pm-skills is installed, then invoke /pm-skills:deliver-prd "test feature" and expect a complete PRD artifact back.
  • Claude.ai / Claude Desktop: ask “Use the prd skill to create requirements for a test feature” and expect a complete PRD artifact back.
  • Other AI agents (Cursor, Windsurf, Copilot, Gemini CLI, and similar): ask “Use the hypothesis skill to test my assumption about checkout abandonment” and expect a structured hypothesis artifact back.

If you get a generic response instead of a structured artifact, the agent has not discovered the skill; see Setup by Platform for troubleshooting.

/pm-skills:deliver-prd "Feature description"
/pm-skills:define-hypothesis "Assumption to test"
/pm-skills:deliver-acceptance-criteria "Story or feature slice"
/pm-skills:deliver-user-stories "PRD or feature context"
/pm-skills:discover-competitive-analysis "Market or product area"

See AGENTS.md for the complete command list.

Skills start cold by default, so each one asks for your phase, your initiative, and anything an earlier skill already produced. Record it once instead.

Add .claude/pm-skills.local.md to your .gitignore first, then create it:

---
schema: 1
phase: discover
active_initiative: "Self-serve onboarding"
---

Eight skills read it. The handoff worth trying: run discover-interview-synthesis, confirm the entry it offers to record, then run deliver-prd and watch it use the personas you already produced instead of asking for them again.

Nothing happens until the file exists, and writes are proposed for your confirmation rather than applied. Full schema and the list of participating skills: Hooks and project memory.

Run multi-skill workflows:

/workflow-feature-kickoff "Feature name" # Problem → Hypothesis → PRD → Stories

Workflow definitions are in _workflows/.

Three utility skills manage the skill library itself:

flowchart LR
    Create["utility-pm-skill-builder\nCreate"] --> Validate["utility-pm-skill-validate\nValidate"]
    Validate -- "PASS" --> Ship["Ship"]
    Validate -- "Findings" --> Iterate["utility-pm-skill-iterate\nIterate"]
    Iterate --> Validate

See the pm-skill-lifecycle guide for detailed workflow patterns.

skills/ # All 68 skill definitions (30 phase + 11 foundation + 12 utility + 15 tool, flat)
commands/ # 11 command markdown files (10 workflow + /chain)
_workflows/ # Multi-skill workflows
scripts/ # sync, validation, and release helpers
.claude/pm-skills-for-claude.md # instructions for Claude Code users
AGENTS.md # Agent discovery index

For Claude Code discovery, run ./scripts/sync-claude.sh (or .ps1) to populate .claude/skills and .claude/commands from the flat source.


Built by Product on Purpose for PMs who ship.