Skip to content

PM Skills - Setup Guide

Source: product-on-purpose/pm-skills via skills.sh - 44.7K combined installs across 71 indexed listings; first seen Oct 9, 2026 (extended-query probe) GitHub: product-on-purpose/pm-skills (716 stars, Apache-2.0; pushed 2026-10-08; layout skills/<name>/SKILL.md) Category: Product Management Quality Tier: 🟡 Beta - 716-star Apache-2.0 library; active Oct 2026; all sampled verdicts Pass

PM Skills is a curated library of 68 plug-and-play product management skills for AI agents, published by product-on-purpose. It covers the complete product lifecycle: 30 Triple Diamond phase skills, 11 foundation skills, 12 utility skills, and 15 tool skills across the Foundation Sprint and Design Sprint workshop families plus note-and-vote. The repo also ships 6 sub-agents, workflow orchestrators, 200+ real-world sample outputs, and CI-enforced contracts.

The library is designed to route between skills: a build-risk review hands off to a hypothesis skill, survey analysis points at experiment design, and so on. The publisher recommends installing the whole set - subsets via --skill work, but handoffs are only seamless with the full library. Skills invoke by name after install, for example /pm-skills:deliver-prd.

Installation

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

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

Cross-agent (skills CLI)

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

The skills CLI scans the skills/ directory and installs the library into your agent's default skills directory. Works with Claude Code, Cursor, GitHub Copilot, Cline, and other agents in the skills ecosystem.

Clone (manual, everything included)

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

Use this path when you want local access to the sample outputs, plan to customize skills, or prefer not to depend on a CLI. Optional: run ./scripts/sync-claude.sh (macOS/Linux) or ./scripts/sync-claude.ps1 (Windows) to populate .claude/skills/ for agents that use that discovery path.

What It Provides

Skill Installs Use For
deliver-acceptance-criteria 1,002 Given/When/Then testable acceptance scenarios
deliver-prd 950 Comprehensive product requirements document with problem, metrics, stories, scope, and dependencies
deliver-user-stories 847 INVEST-compliant user stories with acceptance criteria
deliver-edge-cases 818 Error states, boundaries, and recovery paths for a feature
discover-competitive-analysis 793 Mapping the competitive landscape for gaps and differentiation opportunities
define-problem-statement 767 Crystal-clear problem framing with scope, impact, and success criteria
define-hypothesis 763 Testable assumptions with measurable success conditions
develop-adr 754 Architecture Decision Records in Michael Nygard format
deliver-launch-checklist 744 End-to-end launch checklist so nothing gets missed
define-jtbd-canvas 732 Jobs to be Done framework for customer motivation
develop-solution-brief 730 One-page solution pitch with tradeoffs and open questions
define-opportunity-tree 730 Teresa Torres-style outcome-driven opportunity mapping
discover-interview-synthesis 729 Turning raw user research into insights, patterns, and design implications
measure-dashboard-requirements 720 Analytics dashboard specifications
deliver-release-notes 715 User-facing release communication
measure-experiment-design 714 Rigorous A/B test planning with hypothesis, sample size, and success criteria
measure-instrumentation-spec 709 Event tracking requirements for engineers
develop-design-rationale 709 Documenting why a design choice was made, for future reference
foundation-persona 708 Product or marketing personas with evidence and confidence ratings
iterate-retrospective 706 Team retros that produce real action items, not just feelings
measure-experiment-results 705 Documenting and synthesizing learnings from completed experiments

The remaining 50 indexed listings range from 60 to 699 installs.

Why This Matters for Hermes Agents

Product management is one of the highest-leverage places to point an agent, and one of the easiest places for low-quality output to show: stakeholders who have read hundreds of PRDs notice boilerplate immediately. This library encodes the quality bar. Every skill pairs an instruction set with templates and sample outputs - 200+ real-world examples ship in the repo as the acceptance bar - and the repo's contracts are CI-enforced so skill files stay structurally consistent. The lifecycle structure matters for agent workflows: 30 phase skills map onto Discover, Define, Develop, Deliver, Measure, and Iterate, with foundation, utility, and workshop tool families on top. Skills hand off deliberately (a build-risk review routes into a hypothesis skill; survey analysis points at experiment design), and 10 workflow commands plus a /chain runner orchestrate multi-skill sequences without re-prompting. For teams maintaining their own agent catalogs, the utility family - pm-skill-builder, pm-skill-validate, and pm-skill-auditor - doubles as meta-tooling for authoring and auditing skills.

Usage

You say What happens
Write a PRD for the new onboarding flow deliver-prd drafts problem, metrics, stories, scope, and dependencies
Break this epic into shippable work deliver-user-stories writes INVEST-compliant stories with testable acceptance criteria
Stress-test this feature before we build deliver-edge-cases maps error states, boundaries, and recovery paths
Design the A/B test for our pricing page measure-experiment-design produces hypothesis, sample size, and success criteria; measure-experiment-results reads the outcome later
Synthesize these user interviews discover-interview-synthesis returns actionable insights, patterns, and design implications
Run our sprint retrospective iterate-retrospective ends with real action items, not just discussion
Kick off the new feature end to end the Feature Kickoff workflow chains problem-statement, hypothesis, prd, user-stories, and launch-checklist

Verification

Confirm the skills CLI install with:

npx skills list | grep pm-skills

In Claude Code, invoke a skill directly - /pm-skills:deliver-prd should resolve after a plugin install - and confirm the workflow commands (/workflow-*) are present. Before installing, review the exact instructions the agent will read: the repo holds one folder per skill at skills/<name>/SKILL.md. For example, fetch the acceptance-criteria skill file at https://raw.githubusercontent.com/product-on-purpose/pm-skills/main/skills/deliver-acceptance-criteria/SKILL.md (verified reachable Oct 10, 2026).

Security

skills.sh verdicts for sampled skills (verified Oct 10, 2026) - re-check the security pages on skills.sh before production use:

Skill Gen Agent Trust Hub Socket Snyk
deliver-acceptance-criteria Pass Pass Pass
deliver-prd Pass Pass Pass
deliver-user-stories Pass Pass Pass

Integration with CorpusIQ

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects

Limitations

  • Install counts are flat outside the top of the chart - the bottom 50 listings sit between 60 and 699 installs - so treat popularity as a weak signal for most of the library.
  • Subset installs via --skill can strand handoffs; the publisher recommends the full library or following the routing table to its destinations.
  • Sampled verdicts cover the three deliver-* skills above only; re-check the rest on skills.sh before production use.
  • The index (71 listings) and the documented library (68 skills) do not match one-to-one, so a few indexed entries sit outside the documented set.

  • Nothing runs at install except opt-in hooks, per the project's provenance documentation - review those hooks before enabling them.

  • Snapshot data, verified Oct 10, 2026: 44,697 combined installs across 71 indexed listings; 716 GitHub stars; Apache-2.0; last pushed 2026-10-08. Counts drift over time.

Pro Tips

  • Use the plugin path in Claude Code: /plugin marketplace add product-on-purpose/agent-plugins then /plugin install pm-skills@product-on-purpose gets all 68 skills plus the workflow commands with no clone.
  • Keep project context across sessions with .claude/pm-skills.local.md (gitignored); eight skills read it so you stop re-supplying phase and initiative details.
  • Trust the samples: the repo's 200+ real-world outputs are the acceptance bar for what a good deliverable looks like.
  • Chain instead of re-prompting - workflows such as Feature Kickoff, Lean Startup, and Customer Discovery, or the /chain runner, sequence skills for you.
  • The skills CLI install path notes anonymous telemetry; opt out with DISABLE_TELEMETRY=1 or DO_NOT_TRACK=1.

← Skills Catalog | Skills Marketplace → Powered by CorpusIQ

AI answers on your data. See where every number comes from.

Connect your tools and ask in ChatGPT, Claude, or Perplexity. Enterprise-grade answers in minutes. From $29.95/month.

Start free, no credit card
Contact Terms Privacy Pricing