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 well-crafted, lean diagnostic skill: copy-paste-ready commands, a clear classify-then-fix workflow, useful guardrails against common conflations, and explicit output expectations. The main improvement space is a small amount of concrete interpretation guidance — example codesign/spctl output patterns and a sample repair command — which would lift actionability and workflow closure to top marks.
Suggestions
Actionability: add one-line expected-output cues for the key commands (e.g., what `Authority=adhoc` or `rejected (the code is valid but does not satisfy the specified policy)` from spctl/codesign indicates), so classification is directly grounded in observed output.
Actionability: include a minimal repair example such as `codesign --force --deep --sign - <path>` for the local ad-hoc case, since the workflow promises 'the shortest set of validation or repair commands' but only validation commands are shown.
Workflow clarity: close the loop with an explicit re-validation step after a proposed fix (re-run `codesign -dvvv` / `spctl -a -vv` to confirm the failure class is resolved).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and efficient with zero padding — "Use this skill when the failure smells like codesigning rather than compilation" goes straight to trigger conditions, and no section explains concepts Claude already knows. Every token earns its place, matching the anchor 5 example of terse, well-labeled sections. | 5 / 5 |
Actionability | Concrete, executable commands are given verbatim ("codesign -dvvv --entitlements :- <path>", "spctl -a -vv <path>", "security find-identity -p codesigning -v") plus a concrete seven-class failure taxonomy. It stops short of anchor 5 because there are no example command outputs to interpret and no repair/re-sign command despite the body asking for a "set of validation or repair commands". | 4 / 5 |
Workflow Clarity | A clear four-step sequence (locate binary → read signing details → classify failure → explain minimum fix) with each step's concrete commands. It is not anchor 5 because validation is implicit in the diagnostic commands rather than an explicit validate-and-recover loop (e.g., re-checking after applying a fix), though the operation is diagnostic and non-destructive so no cap applies. | 4 / 5 |
Progressive Disclosure | The skill is a short (~54-line), single-purpose instruction skill with well-organized sections (Quick Start, Workflow, Useful Commands, Guardrails, Output Expectations) and no external bundle files; the body references no missing files (verified: no references/, scripts/, or assets/ directories exist). Per the rubric's simple-skill note, well-organized sections alone earn the top score. | 5 / 5 |
Total | 18 / 20 Passed |