Content
72%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A concise, well-organized checklist with genuinely specific security patterns to check, but it functions as a static list rather than a workflow: there is no ordering of checks, no verification commands, and no instruction for handling discovered violations. Adding a short review sequence with per-finding actions would materially improve it.
Suggestions
Add a brief step sequence for conducting the review (e.g., scan env/config for key exposure, then check schema for RLS, then verify endpoint scoping) instead of presenting an unordered checklist.
Include concrete verification commands or queries, such as SQL to list tables without RLS (e.g., SELECT tablename FROM pg_tables WHERE rowsecurity = false AND schemaname = 'public').
Specify what to do when a check fails — e.g., severity classification (critical for exposed service_role keys vs. warning for missing security_invoker) and whether to fix or only report findings in the PR review record.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~25-line body is lean with no explanations of concepts Claude already knows, but the opening sentence ('Use this skill when reviewing code that touches authentication, database schemas, environment variables, storage, or external API gateways') duplicates the frontmatter description, and the protocol paragraph is peripheral boilerplate that could be trimmed — minor over-explanation matching the 4 anchor. | 4 / 5 |
Actionability | Checklist items name exact patterns to look for ('service_role', 'NEXT_PUBLIC_', 'auth.role()', 'TO authenticated', 'user_id', 'WITH (security_invoker = true)', named buckets), which is concrete and actionable for an instruction-only skill. However, it states what to verify without any commands or queries for how to verify (e.g., no SQL to find tables lacking RLS), leaving minor gaps consistent with the 4 anchor rather than fully copy-paste-ready guidance. | 4 / 5 |
Workflow Clarity | The protocol section gives rough before/during/after sequencing ('Follow `AGENTS.md`... before starting', 'After a completed PR review, post one `Review record` comment'), but the review checklist itself is an unordered list with no explicit sequence, no validation checkpoints, and no guidance on what to do when a violation is found (report vs. fix). This matches the 3 anchor (steps present, checkpoints missing) and is below the 4 anchor's 'most checkpoints present'. | 3 / 5 |
Progressive Disclosure | The skill is under 50 lines with no bundle files and no need for external references; its headed sections ('Repository Review Protocol', 'Key Confinement & Environment Safety', 'Database & Storage Safety') are well organized, and repo-doc references are one level deep and clearly signaled, so the simple-skill exception applies for a top score. | 5 / 5 |
Total | 16 / 20 Passed |