Content
77%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 is a strongly actionable, well-sequenced audit workflow with concrete greps, verification steps, fixes, and a structured output format — top marks for actionability and workflow clarity. Its weaknesses are token efficiency (coaching and example sections add padding) and monolithic structure (the 10-pattern catalog would be better split into reference files).
Suggestions
Trim the 'Invocation Examples' section (five near-identical examples) and the 'Scope' section, which restates the description, to reclaim tokens without losing audit capability.
Move the 10-pattern detection catalog (Phase 2) into a references/ file (e.g. references/patterns.md) and keep a one-line summary table in SKILL.md, so the overview loads lean and detail is one level deep.
Fold the 'Tool Use Is Mandatory' section's guidance into a single line at the top instead of a dedicated section with repeated instructions.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The pattern catalog is dense and operational (Issue/Search/Verify/Fix per pattern), but sections like 'Invocation Examples' (five near-identical examples), 'Scope' (restates the description), and 'Tool Use Is Mandatory' (coaching model behavior) are padding that could be trimmed. Not score 4 because several sections add tokens without adding audit capability; not score 2 since the bulk is genuinely operational. | 3 / 5 |
Actionability | Every pattern ships a runnable grep regex, a Read-verification step, and copy-paste fixes (e.g. 'try db.write { db in for item in items { try item.insert(db) } }' and the 'db.columns(in:)' existence guard), plus a complete output template and a false-positive list. This is fully executable guidance covering the common cases. | 5 / 5 |
Workflow Clarity | Five clearly sequenced phases (map → detect → completeness → cross-reference → score) with explicit validation checkpoints: 'use Read to verify the surrounding context before reporting', the Schema Map gate before Phase 2, false-positive checks, and output limits. As a read-only audit there are no destructive operations, so the batch/destruction cap does not apply. | 5 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ are absent), and the ~290-line body inlines the entire 10-pattern catalog — exactly the bulk detail a references file would hold. Headers are clear, giving structure, but content that should be separate is inline and the 'Related' links point at another skill's files rather than a local bundle. Not score 4 because the bulk detail is not split out. | 3 / 5 |
Total | 16 / 20 Passed |