Skills
Open-source Claude Skills that package how I actually work as a product manager — download the markdown, drop it into Claude, and run it yourself.
Not generic templates. Each skill encodes an opinion the model wouldn't otherwise hold — rules earned the hard way, so every output comes out the way it should.
The Skills Library
The competencies above, packaged so an AI agent can actually run them. Each file is a Claude Skill — a set of instructions Claude loads on demand when the work matches. They carry opinions, not templates: what a PRD must never ship without, how a discovery gets stress-tested, what a job application is not allowed to claim.
prd-writer
Writes and reviews PRDs, one-pagers, and feature briefs. Problem before solution, mandatory non-goals, the riskiest assumption named out loud, confidence-flagged numbers, and vertical slices instead of layers.
discovery-plan
Turns a fuzzy space into a defensible point of view. Attacks the problem set, runs a stakeholder-as-agents workshop in parallel, insists on one real practitioner conversation, and keeps an audit trail of what got dropped and why.
job-application
Tailors a resume, cover letter, or application answers to one JD. Reads the posting as a product spec, scores the fit honestly including the gaps, picks three proof points, and refuses to invent anything.
How to use one
Drop the downloaded file into a folder named after the skill, inside your Claude skills directory. Claude reads the description and loads the skill on its own when your request matches — there's nothing to invoke.
~/.claude/skills/prd-writer/SKILL.mdWorks in Claude Code and any agent that reads skill files. Free to take, edit, and make your own — the opinions in them are mine, and you should overwrite them with yours.
Common questions
How does AI change the PM skill stack? expand_more
AI automates the "glue" work — PRDs, ticket triage, status syncs — and rewards higher-order skills: clear problem definition, ethical guardrails, prompt engineering, and orchestrating AI agents alongside human teammates. See the Agent Lab for concrete examples.
What is a Claude Skill, and how is it different from a prompt? expand_more
A prompt is something you retype every time. A skill is a markdown file the agent loads by itself when the work matches — it has a name, a description that acts as the trigger, and the instructions underneath. The practical difference is consistency: every PRD comes out with non-goals in it because the skill says so, not because you remembered to ask.
Why write your own skills instead of using generic PM templates? expand_more
A skill earns its place only if it encodes an opinion the model wouldn't otherwise hold. Reproducing a public PRD template adds nothing — Claude already knows it. The value is in the rules you've earned the hard way: always name the riskiest assumption, never estimate in story points, default AI bets to a foundation model before a custom one, flag every number by confidence.
Ready to Collaborate?
Let's deploy these skills to build something that actually shifts the needle. No fluff, just hard results.