Content
93%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.
An excellent routing-skill body: lean, executable recon commands, a well-designed sub-skill routing table with explicit load conditions, and a strong safety decision gate. The only weakness is that the recon workflow is conveyed through section ordering rather than an explicit step sequence or feedback loop.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 110-line body is lean and dense: a routing table, provider-fingerprint commands, a risky-surface triage table, grep patterns, and a tool table. Every token earns its place; there is no padding and no explanation of concepts Claude already knows. | 5 / 5 |
Actionability | Fully executable guidance throughout: copy-paste-ready 'gh api', 'grep -rE', 'find', and 'test -f' commands with concrete regexes, plus named tools (gato, actionlint, zizmor) with their specific uses. The risky-surface table pairs each signal with an exact grep. | 5 / 5 |
Workflow Clarity | A logical recon sequence is present (fingerprint provider → enumerate workflow files → triage risky surface → check secret exposure) and the decision gate provides an explicit validation checklist before active testing, so the destructive-operation cap does not apply. However, the sequence is implied by section order rather than explicitly enumerated, and there are no error-recovery feedback loops, which keeps it below a 5. | 4 / 5 |
Progressive Disclosure | This is a routing skill that does progressive disclosure correctly: the sub-skills table pairs each leaf with a 'When to Load' condition and an explicit load_skill() path, and all attack detail is deferred one level deep to the leaves. Core recon commands are appropriately kept inline; no bundle files exist locally, and the leaf references are clearly signaled external paths. | 5 / 5 |
Total | 19 / 20 Passed |