PluginPartly tested (needs an account)

Ship a product with agents, no coding

Product Engineering

Free MIT skill packs that guide you from an idea to a paid product: proposal, build, independent review, release check and running it with real users.

Works with
Claude Code, Codex CLI, Gemini CLI
Good for
Build an app, Web developers

What it does

Product Engineering is a set of six skill packs (44 skills) for people who build products with coding agents but do not write the code themselves. You make the decisions and check the result; the skills give the assistant the method.

It covers eight steps: say what you have, review the proposal (one HTML page with the research, the key screens and the open decisions), plan, build, click through it, review the release, release and learn from real users. Along the way there are skills for competitor and market research, a fresh reviewer that checks the code before each release, go or no-go release checks, and production setup for payments, sign-in, email and monitoring.

If you get lost, /product:start-here tells you where the work stands and what to do next. Small changes go straight to the build, and only bigger ones get a written plan.

The six packs are: product (deciding what to build), ai-shipping (review and release), agent-teams (working with other coding agents), agent-harness (a reliable agent setup), app-stack (a SaaS on Next.js, Supabase, Clerk and Netlify) and work-humanizer (work writing that does not read as AI). You can install only the packs you need: the first two lines are the core.

Example prompts

  • "/product:start-here"
  • "I have an idea for a booking app for my salon. Propose what to build first."
  • "/ai-shipping:ship-check"
  • "/agent-harness:handover"

How to install

claude plugin marketplace add phuryn/product-engineering
claude plugin install product@product-engineering
claude plugin install ai-shipping@product-engineering
claude plugin install agent-teams@product-engineering
claude plugin install agent-harness@product-engineering
claude plugin install app-stack@product-engineering
claude plugin install work-humanizer@product-engineering

How to use it

Type /product:start-here and say what you have, for example "I have an idea for a booking app for my salon", and it tells you where the work stands and what's next.

Check it works

After installing, start Claude Code in your project (run /reload-plugins if it is already open), then type /product:start-here. It should ask what you have and suggest the next step.

Common problems

Other assistants

What to watch out for

  • The README's quickest install is a text to paste to your assistant ("Please install skills from phuryn/product-engineering ... Enable auto updates"). That lets the assistant pick the steps itself and turn on automatic updates; use the exact commands above instead
  • The independent-review and ship-check skills run a Python wrapper (scripts/review.py) that starts a second assistant as the reviewer: the Grok CLI by default when it is installed, so your code goes to xAI; otherwise a restricted claude -p session. The agent-teams skills drive the Codex CLI, the Grok CLI or Claude Code headless. Each of these uses that tool's own account, plan or credits
  • The app-stack skills set up real production services (Supabase, Clerk, Netlify, payments, email, monitoring). The uptime-monitoring skill proposes Staygreen by default, a service made by the author (its SKILL.md says so), and can walk you through making an account; check each step before you approve it
  • The guardrail-hooks skill writes hooks into your agent setup, and agent-instructions edits AGENTS.md or CLAUDE.md, when you ask for them. The packs themselves add no hooks, sub-agents or MCP servers
  • guardrail-hooks' one-way-door guard does not catch everything. Its SKILL.md says it does not see python -c or node -e code, a script run after a cd, ./script, python -m, uv run x.py, piped code or a path with spaces, "so a delete reached that way runs unasked". If you installed its scripts before 1.8.5, copy the new guard_one_way_doors.py and test_hooks.py over your copies and keep your rules. Test it only with the safe canary true guard-canary, never with a real delete or push (releases 1.8.5 and 1.8.6)
  • The scripts need Python 3.9 or newer; nothing runs unless your assistant runs it. The npx skills route (other agents) needs Node.js 22.20 or newer, and the skills installer sends anonymous install data unless you set DISABLE_TELEMETRY=1
  • The work-humanizer pack is a newer copy of the separate work-humanizer catalog item (the SKILL.md files differ a little). Keep one: if you already installed work-humanizer on its own, skip the last install line, or the skill loads twice
  • Since 1.7.0 the design skill is called ui-design (in app-stack). If your CLAUDE.md or AGENTS.md says "apply design", change it to "apply ui-design", and delete an old copy in .claude/skills/design/. propose now writes one page, docs/product/proposal.html, and the next run replaces an old docs/product/proposal.md with it (git keeps the history)
  • With auto-update on, a push to the main branch reaches your install even without a release. The 1.8.0 notes say two unreviewed commits, one of them the 1.7.0 rename, reached auto-updated installs within the hour. The author now works on a separate branch and moves main only at a release
  • Five product skills are also in the author's pm-skills collection; with both installed each loads twice. The README says to turn off pm-ai-shipping once ai-shipping is installed

Licence and author

Licence
MIT
Author
Paweł Huryn (phuryn)
Source code
github.com/phuryn/product-engineering

How we tested it

We installed it on an empty test computer on 3 Nov 2026. Partly tested: installed in a clean test system, but its main features need an account or key we did not use.

Test level
Technical details of the test
Date
3 Nov 2026
Method
Reviewed by our team against the repository, its README and docs.

Checked again on 9 Nov 2026

Alternatives

  • work-humanizer — the same author's writing skill on its own; it is also the sixth pack here, so install it only once
  • Superpowers — general build workflows without the product and business steps
  • PR Review Toolkit (Anthropic) — only the code review part
  • Conductor — plans features as specs before coding

FAQ

  • Q: Do I need to code? A: No. The README says it is for people who make decisions and check the product, not every line of code.
  • Q: Does it cost money? A: The skills are free (MIT). You pay for your assistant plan, any second assistant used as a reviewer (Grok or Codex) and any services you connect.
  • Q: Do I need all six packs? A: No. Install only the packs you need.
  • Q: I already have work-humanizer. Should I install it again? A: No. Skip claude plugin install work-humanizer@product-engineering, or remove your separate copy. Two copies of the same skill load twice.

Sources

  • github.com/phuryn/product-engineering
  • github.com/phuryn/product-engineering/tree/67a8e4b3fd64bfebeb86819af90bc459cd6f6d6e
  • github.com/phuryn/product-engineering/blob/67a8e4b3fd64bfebeb86819af90bc459cd6f6d6e/README.md
  • github.com/phuryn/product-engineering/blob/67a8e4b3fd64bfebeb86819af90bc459cd6f6d6e/.claude-plugin/marketplace.json
  • github.com/phuryn/product-engineering/blob/67a8e4b3fd64bfebeb86819af90bc459cd6f6d6e/ai-shipping/skills/independent-review/SKILL.md
  • github.com/phuryn/product-engineering/blob/67a8e4b3fd64bfebeb86819af90bc459cd6f6d6e/app-stack/skills/uptime-monitoring/SKILL.md
  • github.com/phuryn/product-engineering/tree/67a8e4b3fd64bfebeb86819af90bc459cd6f6d6e/work-humanizer
  • github.com/phuryn/product-engineering/releases/tag/v1.7.0
  • github.com/phuryn/product-engineering/releases/tag/v1.8.0
  • github.com/phuryn/product-engineering/releases/tag/v1.8.5
  • github.com/phuryn/product-engineering/releases/tag/v1.8.6
  • github.com/phuryn/product-engineering/blob/063a34716c18b5c939214cfe90a2736822706710/agent-harness/skills/guardrail-hooks/SKILL.md
  • learn.chatgpt.com/docs/cli/reference
  • github.com/vercel-labs/skills