# Gsd Loop Review

> Audit one open PR against its linked issue contract and required CI, then post a gsd-loop verdict and labels. Use when asked to run the reviewer or audit the PR queue. Never merge or push; each invocation completes one pass.

- Skill: `open-gsd/gsd-loop-review` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx skillmds@latest add open-gsd/gsd-loop-review`
- Raw SKILL.md: https://api.skillmd.com/api/skills/open-gsd/gsd-loop-review/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: open-gsd (https://skillmd.com/u/open-gsd)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/open-gsd/gsd-loop-review

---


Resolve the repository root with `git rev-parse --show-toplevel`. Use its
`loop/review.md` only when the root identifies itself as the gsd-loop source:
`README.md` starts with `# gsd-loop`, all four files under `loop/` exist, and
`scripts/doctor.sh` plus `scripts/scheduler-policy.sh` exist. Otherwise, read
`playbook.md` beside this `SKILL.md`; global installs bundle that canonical
fallback.

Before executing, resolve `OUTCOME_SYNC` to `scripts/sync-outcomes.mjs` beside
this `SKILL.md`. Pass that absolute path wherever the playbook says
`OUTCOME_SYNC`; installed skills bundle its runtime, so a global `gsd-loop`
command is not required. Resolve `AUDIT_VALIDATOR` the same way to
`scripts/validate-audit-evidence.mjs`. Execute exactly one pass.