| name | paul-review |
| description | Run thorough code review via REV subagent — spawns an isolated reviewer with fresh context for unbiased, senior-dev-level review across 8 dimensions. |
PAUL Review
Run an on-demand REV review by dispatching an isolated code-reviewer subagent.
Canonical references
../references/review-prompt.md
../references/review-checklist.md
Default output
- Concise by default: show scope, reviewer dispatch status, findings summary, verdict, and one next action.
- Expand for critical/high findings, unavailable subagent support, oversized diffs, degraded review mode, or user request.
Command-local notes
- Read
.paul/STATE.md and pals.json (modules.rev.enabled, model, pr_review, blocking settings).
- Ask for one review scope: current plan changes, specific files, branch vs base, or uncommitted changes; map scope using
review-prompt.md.
- Assemble diff, changed files, optional
AGENTS.md, and related *.test.* / *.spec.* files.
- Dispatch
Agent({ subagent_type: "code-reviewer", run_in_background: false }); pass modules.rev.model when configured.
- Review dimensions stay in the referenced checklist: correctness, edge cases, naming, architecture, performance, security, test coverage, rollback safety.
- If
Agent is unavailable, offer skip or degraded in-session review; label degraded mode clearly.
- Preserve structured findings with severity, location, suggestion, strengths, and verdict:
READY, READY WITH CONCERNS, or NOT READY.
- On-demand only — NOT a lifecycle hook; REV does not fire automatically during PLAN/APPLY/UNIFY.