Review Best Practice Skill

Evaluate a proposed SKILL.md against grimoire standards and produce a structured PASS / NEEDS-REVISION / REJECT verdict with specific findings per criterion.

Why This Is Best Practice

Adopted by: Deterministic review checklists are standard in high-quality OSS projects — Linux kernel patch review, CPython PEPs, React RFC process, and most projects with a quality bar use structured, criteria-based review over subjective human judgement. Impact: Structured checklist-based review (Fagan Inspection method, IBM 1976) achieves 60–90% defect detection — highest of any quality technique per Capers Jones (Software Quality Engineering, 2011). Consistent criteria reduce reviewer disagreement from 40–60% (ad-hoc review) to <10% (checklist-based) per Bacchelli & Bird, ICSE 2013. Agent execution reduces review turnaround from days to seconds. Why best: A checklist applied by humans drifts over time and varies by reviewer. A skill applied by an AI agent produces identical criteria evaluation every time, making the standard self-enforcing rather than aspirational.