Content
86%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 tightly written, well-sequenced security review process that assumes Claude's competence and adds only project-specific guidance. The main gap is the absence of any concrete tooling, search commands, or a worked example for the search-and-prove steps, and no explicit recovery path for unproven findings.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is six lean numbered lines with zero padding and no explanation of concepts Claude already knows; every phrase ("Rank by impact and likelihood; avoid speculative style findings") adds non-obvious guidance. This matches the lean-and-efficient anchor where every token earns its place, not the level below, which would require trimming minor over-explanation. | 5 / 5 |
Actionability | The guidance is concrete for an instruction-only skill: it enumerates specific vulnerability classes to search ("auth, injection, SSRF, file handling, secrets, storage, RLS, and dependencies"), specific evidence standards ("changed code, a safe local test, or deterministic static evidence"), and a defined reporting taxonomy. It stays at mostly-executable rather than fully actionable because no commands, search patterns, or worked examples are given for steps like "Search for realistic exploit paths". | 4 / 5 |
Workflow Clarity | A clear six-step sequence is present with an explicit validation checkpoint ("Prove each finding with changed code, a safe local test, or deterministic static evidence") and safety gates (fixes only when requested, live validation approval-gated). It falls short of the top anchor because there is no feedback loop for error recovery, e.g., what to do when a suspected finding cannot be proven, though step 4's "avoid speculative style findings" implicitly handles it. | 4 / 5 |
Progressive Disclosure | The skill is under 50 lines, has no external references to organize, and no bundle files exist; the single well-ordered process list under one heading is the appropriate structure. Per the rubric's simple-skill guidance, this earns the top score without external file references. | 5 / 5 |
Total | 18 / 20 Passed |