aw-review-quality-gate

# Review Quality Gate Run this gate on every finding you have generated, **before** formatting output or delivering results. The gate catches false positives, noise, and miscalibrated severity — the most common failure modes of single-pass review. --- ## When to Use - As the `autonomous-workflow` Phase 6 review companion — invoked via `Skill()` after `review-changes` returns findings, before the workflow acts on them. See [`phase-6-pr-creation.md#findings-quality-gate`](../autonomous-workflow/rules/phase-6-pr-creation.md#findings-quality-gate). Graceful skip: the workflow acts on the raw findings if this skill is not installed. - After the `pr-reviewer` agent generates findings (Step 2.5). - After DX or UX review skills produce their finding lists. - Any time a skill produces actionable recommendations that a human will read. --- ## Gate Checklist For **each finding**, answer every question. Be honest — a dropped false positive is worth more than a delivered one. | # | Question | Fail means | |---|----------|------------| | 1 | **Is it actionable?** The author can do something concrete. No "consider", "might want to", or "could be improved" without a specific fix. | Finding is vague noise | | 2 | **Did I read the surrounding code?** Not just the diff hunk — the full function, the caller, the guard clause three lines up. | Finding may be wrong | | 3 | **Is severity calibrated?** A style nit is never "required." A SQL injection is never "suggestion." Re-check the severi...