build-review
SDD review + dogfood skill (v3), two modes by explicit arg. **pipeline-review** (the /build phase gate): Round 1 code r…
Description
SDD review + dogfood skill (v3), two modes by explicit arg. **pipeline-review** (the /build phase gate): Round 1 code review — automated checks (tsc, unit tests, one curl per API contract), then the shared `code-reviewer` agent (all architecture/Fowler/ponytail lenses + the correctness bug-hunt, one pass) — re-verified, no browser; Round 2 dogfood through the shared `dogfood` agent (blind first-impression + guided story/shell/validation walk in one session), then outcome-card grading; every HIGH+MEDIUM finding auto-fixes silently in one wave-dispatched loop (cap 3 → Accept/Stop binary); then it owns the dogfood handoff and writes phase-complete | phase-blocked. **standalone-dogfood** (NON-pipeline): auto-fires whenever implementation finishes OUTSIDE /build — "done", "just built X", "just fixed X" — scoped to the just-built feature, runs the SAME two agents + fix loop, writes NOTHING to build state and needs no .build-state.json or spec dir. Trigger on /build-review, when /build reaches its review step, or wh. A reusable SKILL.md agent skill by anhtrinh919.
Author
anhtrinh919
Platform
cli
Pricing model
free
Categories
Tags
Capabilities
- SKILL.md package