phuryn/product-engineering

What it is

Product Engineering is a collection of 43 agent skills (six packs: product, ai-shipping, agent-teams, agent-harness, app-stack, work-humanizer) for taking a software product from the first idea to paying users. The intended users are product managers, founders and designers who ship with coding agents and review the product rather than every line of code, plus engineers responsible for the whole product, not just the code (who it's for).

The skills run in Claude Code, Codex, Cursor, Gemini CLI, OpenCode, GitHub Copilot and any agent that reads SKILL.md folders. The method has three phases — discover what to build, build and check, ship and learn — mapped to eight steps from "Start" to "Real users" (process overview). Evidence comes first: the agent researches the alternatives and segment, proposes a direction with sources, builds a prototype on sample data, and asks you to approve it. During the build, one session coordinates, another builds, and a fresh reviewer that never saw the build reviews every change. A release gate collects evidence before promoting, and production concerns (authorization, billing, email, analytics, monitoring) are part of the collection rather than left later.

The companion PM Skills collection covers discovery, strategy and go-to-market; this one covers building, shipping and running the product. The ai-shipping pack moved here from pm-skills.

Quick start

Install into a project with your agent. The simplest route is to open the agent in the project folder and ask it to install from phuryn/product-engineering, enabling auto updates (install instructions). If your agent can't install from a prompt, use the explicit commands.

Claude Code, user scope, then enable auto-update in /plugin → Marketplaces → product-engineering (commands):

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

Add --scope project to keep the packs to the repository, or --scope local to keep them only for you. Codex reads the same marketplace:

codex plugin marketplace add phuryn/product-engineering
codex plugin add product@product-engineering   # and the other packs

For any other agent, use the skills CLI:

npx skills add phuryn/product-engineering          # pick skills and agents interactively
npx skills update                                  # later, to update

After installing, start a session and type /product:start-here in Claude Code or $product:start-here in Codex, then say what you have: an idea, user evidence, an agreed change, a bug or a release. In a terminal, if the command isn't found, run /reload-plugins; in Codex, start a new session first. From there, the start-here skill reports where the work stands and what's next, and the matching skills take over. WALKTHROUGH.md shows the full loop from the first prompt through the next round.

Constraints and practical notes

• The npx skills route requires Node.js 22.20 or later; on older Node it fails at start with a missing glob export. Upgrade Node or run it through npx -p node@22 -p skills -c "skills add phuryn/product-engineering" (install notes).
• Third-party marketplaces don't auto-update by default; without it, update with /plugin → Marketplaces → product-engineering → Update marketplace, or claude plugin update <pack>@product-engineering. Codex updates with codex plugin marketplace upgrade product-engineering followed by a new session.
• Skills that ship scripts require Python 3.10+ and the CLI they drive; nothing runs unless your agent runs it (scripts).
• An agent sees only what is installed for it or committed in the repository it works in. A Claude Code plugin is invisible to Codex in the same folder. For products built with multiple agents, copy the needed skills into the product's .claude/skills/ or .agents/skills/ and commit them; then disable the copied packs per repository to avoid double loading (multi-agent setup).
• Installing the packs adds roughly 2,600 tokens per turn in Claude Code; self-starting skill descriptions are capped at 6,000 characters total by tools/validate.py (cost notes).
• An app-stack SaaS is opinionated toward Next.js, Supabase, Clerk and Netlify.
• When both this collection and pm-skills are installed, disable the old pm-ai-shipping plugin and keep only one copy of work-humanizer; five strategy skills appear in both collections (pm-skills overlap).
• Licensed under MIT (LICENSE.md).

HotRepo's take

The repository was recorded once on HotRepo's new-project list for the Python language board, on 2026-11-01, with 39 stars added over the roughly 12-hour window from 2026-10-31 21:45 UTC to 2026-11-01 09:45 UTC, reaching 140 stars and 22 forks. No GitHub daily rank was recorded, so the data supports only that early interest, not a trend or broader adoption. What the README makes concrete is a deliberate split from engineer-first skill collections: the spec is a clickable product, review is done by a fresh model family, and production concerns are part of the skill set — a structure aimed at people who ship with agents without reading the code.