Intake Repo
Use this workflow to quickly understand a repository before implementation.
Goal
Produce a short, evidence-based repository profile with risks and next steps.
Output
.ai/PROJECT_PROFILE.mdusingtemplates/PROJECT_PROFILE.md.
Rules
- Do not edit product code.
- Do not rewrite architecture.
- Prefer observed code patterns over assumptions.
- Keep profile concise and actionable.
- "Not applicable" is a valid, expected answer for a Main Area — a static/portfolio site with no backend has no auth, database, or AI-logic lane to fill in. Mark it as Not Applicable rather than inventing content, guessing at a future need, or leaving it blank/ambiguous.
Steps
- Detect framework, language, package manager, and deployment target.
- Read package/lock/config files and major entry points.
- Identify available scripts: dev, typecheck, lint, test, build.
- Map frontend, backend, database, auth, AI, deployment areas.
- Identify environment variable usage patterns.
- Capture current architecture patterns used in the code.
- List obvious missing production basics and key risks.
- Generate or update
.ai/PROJECT_PROFILE.md.
Minimum Evidence
Cite concrete files for each finding. Example categories:
- script source (
package.json,pyproject.toml, etc.) - deployment source (
vercel.json, Docker, CI workflows) - env usage source (
process.env,NEXT_PUBLIC_, etc.) - auth/data boundaries (middleware, route handlers, policy files)
Done Check
- Profile includes detected stack, scripts, main areas, patterns, risks.
- Output is short enough to be useful during normal coding tasks.