What problem does it solve?
Prevents risky or inconsistent coding changes by forcing every development task into a clear XS/M/L tier with required RFC gating, an explicit “image of result” approval step, mandatory smoke tests, and multi-agent review (including security review for bigger or auth/data-impacting work).
Core Features & Use Cases
- Tiered development workflow (XS / M / L) that defines ceremony level per task size and risk, including “choose higher if unsure.”
- User-level “image of result” approval before code (Step 0.5) to align UX outcomes with the user before implementation details.
- RFC gating for M/L using
PROJECT/rfc/NNN-title.md to lock the plan (problem → options → chosen → risks → verification) before any coding begins.
- Mandatory smoke test even for XS plus test guidance (unit vs integration vs E2E) to catch refactor breakage quickly.
- Multi-agent review pipeline where M tier uses a code-review subagent and L tier additionally requires
/security-review and /review, culminating in a final user approval for M/L.
- Incremental security levels (S1 → S2 → S3) with different security/scanning/threat-model expectations as the project matures.
- Blameless incident logging and Definition of Done (DOD) via
PROJECT/incidents.md and a pre-report checklist to keep changes consistent and auditable.
Quick Start
Tell the AI: “Dev workflow—classify this coding task as XS/M/L, draft the RFC and image of result if required, then wait for my approval before writing code.”