Content
46%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.
The body has a sound skeleton (real commands, a genuine one-level-deep reference bundle, and clear script parameters) but is buried under heavy governance boilerplate that duplicates itself across many sections and inflates token cost for a simple 37-line script. The core execution guidance is also incomplete: the documented usage example does not match the script's actual text-vs-file behavior.
Suggestions
Cut the boilerplate sections (Risk Assessment, Security Checklist, Evaluation Criteria, Lifecycle Status, Response Template, Output Contract, Failure Handling, User Checkpoints) — they restate the same rules already in Error Handling and could shrink the file by more than half without losing skill-specific guidance.
Fix the usage documentation to match script behavior: main.py treats --manuscript as literal text, not a file path, so either document 'python scripts/main.py --manuscript "...Figure 3 shows..."' or update the script to read files.
Turn the abstract Workflow steps into a concrete executable sequence (run py_compile check, run the checker, report found references vs labels), and specify what the script's output looks like.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is ~190 lines for a 37-line script, with many padded boilerplate sections (Risk Assessment table, Security Checklist, Evaluation Criteria, Lifecycle Status, Response Template, Output Contract, User Checkpoints) that add no skill-specific knowledge. The same error-handling rules ('state exactly which fields are missing', 'do not fabricate') are restated across Error Handling, Failure Handling, Input Validation, and Output Contract, matching anchor 2 (several unnecessary/padded sections) rather than 1, since no basic concepts are explained. | 2 / 5 |
Actionability | Concrete commands exist ('python -m py_compile scripts/main.py', 'python scripts/main.py --manuscript paper.docx'), but key details are missing or wrong: the bundled script treats --manuscript as literal text rather than reading a file, so the documented usage example would print nothing useful; output format is never specified; and the Features (orphaned references, label consistency, missing citations) are not connected to any command. This matches anchor 3 (some concrete guidance but incomplete, missing key details) rather than 4. | 3 / 5 |
Workflow Clarity | The Workflow section lists a sequence and the Quick Check provides one real validation checkpoint (py_compile before deeper execution), but the steps are abstract governance prose ('Confirm the user objective', 'Validate that the request matches the documented scope') with no concrete run-parse-report sequence for the actual script. This matches anchor 3 (sequence present, checkpoints implicit/missing) — the read-only nature of the skill avoids the destructive-operation cap. | 3 / 5 |
Progressive Disclosure | The bundle structure is good: scripts/main.py and references/audit-reference.md both exist, the reference is linked once with a clear description ('Supported scope, audit commands, and fallback boundaries'), and references are exactly one level deep with no nested indirection. This matches anchor 4 (good structure, references mostly clear, minor organization gaps) — it falls short of 5 because large amounts of boilerplate (security checklists, risk tables, response templates) are inlined in SKILL.md rather than split out or removed, slightly bloating the overview. | 4 / 5 |
Total | 12 / 20 Passed |