Product Engineering

Your agent can take a raw product idea to a build-ready decision brief, then hand each piece to your build loop.

Install

You need agentbundle first. Install it once for your user account:

agentbundle install --pack product-engineering

In Claude Code you can install it as a plugin instead:

claude plugin install product-engineering@agent-ready-repo

What your agent can do

• Frame a rough idea as an intent: the problem, who has it, the outcome, and the opportunity behind it.
• Sort a market shift, metric drop, or recurring user pain before anyone shapes work.
• Find and rank the unmet jobs behind a problem, and plan how to validate the riskiest assumptions.
• Lay out solution options with trade-offs, then record the bet with its reasons and what would reverse it.
• Map the capabilities a bet needs and decide build, buy, partner, or adopt for each.
• Break an intent into delivery briefs and specs, and write UI copy for each screen state.

Skills

• discovery-loop — Runs a supervised discovery from idea to decision brief, and resumes where you left off.
• intent — Frames, de-risks, and decomposes product work in three modes.
• frame-situation — Classifies a signal and rates capability maturity on a Wardley map.
• frame-domain — Frames how the activity is really done and draws the MVP boundary.
• identify-opportunities — Ranks unmet jobs by importance and satisfaction.
• plan-validation — Plans what would confirm or kill each assumption, with an interview or test guide.
• explore-options — Sketches several distinct product shapes before you pick one.
• diverge-solutions — Compares solution options for an initiative, with a recommendation.
• lean-canvas — Fills a lean canvas and writes a one-page initiative brief.
• place-bet — Records the choice, confidence, appetite, accepted risks, and reversal conditions.
• map-capabilities — Lists needed capabilities and a build order. On the longer route, intent decompose takes the capability map next.
• align-value-stream — Sets up a coordinating repo for work that spans many components. It adds a hard-limits README section and offers architect-design to write the solution reference architecture.

Subagents: discovery-lead supervises the loop. discovery-reliability-reviewer and discovery-threat-reviewer review discovery artifacts read-only.

Works with

Works with Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, and Kiro IDE. Product Strategy supplies goals and opportunity assessments as anchors. Core takes each result as a delivery contract or brief, and Core is optional. Experience Design owns the screen flows and their UI copy. When the architect pack is installed, frame-domain and intent de-risk offer architect-assess, intent frame and map-capabilities offer architect-design, and intent de-risk and intent decompose offer architect-design before the work is cut. Without it, they say nothing about it. Guides: Shape an idea and Run a discovery.

Questions

Where do I start? Type intent frame and describe a product problem at any level of clarity. To resume, type discovery-loop with the initiative name.

Does it write specs? It decomposes an intent into briefs and specs, and after you confirm, hands each independently shippable piece to Core. Core’s approval path still applies.

Where are the files saved? In docs/product/ and docs/discovery/ by default. Both are configurable in agentbundle-layout.toml. The docs/product/voice/ folder holds voice charts, which Experience Design’s ux-writing writes.

Does it need Product Strategy first? No. Without strategy files, the skills carry on.

Details

Scope: user (default), repo · Version: 2.0.2 · Source: packs/product-engineering