Built as working context, not shelfware.
This resource is meant to be useful inside the tools where product work now happens: your codebase, your notes, and your AI-assisted workflow.
01
Paste the markdown into Claude, ChatGPT, Cursor, Codex, Gemini, or another AI agent as reusable project context.
02
Use it before a planning, implementation, review, or audit session so the agent has constraints, criteria, and working structure up front.
03
Adapt the sections to your product, team, or repo before asking the agent to execute against it.
Markdown previewproduct-spec-agent-template.md
# Product Spec Agent Template Paste this into an AI agent before asking it to plan, critique, or implement a product change. ## Context ```md Product: Surface: Audience: Current behavior: Desired behavior: Primary user job: Business reason: Deadline: Non-goals: ``` ## Constraints ```md Design system constraints: Engineering constraints: Data/API constraints: Permission constraints: Analytics constraints: Accessibility requirements: Mobile requirements: Dependencies: ``` ## Required States - [ ] Loading - [ ] Empty - [ ] Error - [ ] Partial data - [ ] Permission denied - [ ] Long content - [ ] Small viewport - [ ] Slow network - [ ] Success - [ ] Undo or recovery ## Agent Instructions Ask the agent to: 1. Restate the product job. 2. List assumptions. 3. Identify risky missing states. 4. Propose the smallest useful shipped version. 5. Describe implementation boundaries. 6. Write acceptance criteria. 7. Suggest tests and manual QA steps. 8. Flag any invented behavior before implementation. ## Acceptance Criteria ```md User can: System should: System should not: Analytics should capture: QA should verify: Known tradeoffs: ```