Content
85%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 an unusually actionable and well-sequenced verification workflow: exact commands, explicit validation checkpoints, error-recovery loops, and rigid triage rules for every check. Its main weaknesses are the monolithic single-file layout with dangling script references and occasional over-explanation of concepts Claude already knows.
Suggestions
Move the Layer 3/4 implementation detail (CRAP join rules, mutation-testing blind spots, the crap-report field rules) into reference files under references/ and keep one-line pointers in SKILL.md, so the main file stays a navigable overview.
Ship the referenced scripts (scripts/crap-report.bb, scripts/complexity-gate.bb, scripts/quality-gate.bb) in the bundle or remove/qualify the citations, since the commands currently point at files the skill does not provide.
Trim explanations of Rust fundamentals ('forbid (unlike deny)', 'Prefer NonZero*', enums-over-booleans) down to bare reminders, and tighten the greenfield 'how do you know it's right' digression to the pattern ranking itself.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly dense, field-earned guidance (the tmpfs overflow failure mode, fn-hash snapshot protection, line-range-overlap join) where every token earns its place, with only minor over-explanation of Rust basics ('forbid (unlike deny) cannot be overridden', 'Prefer NonZero*') and a somewhat editorial greenfield digression that could be trimmed. | 4 / 5 |
Actionability | Fully executable throughout: copy-paste commands ('export TMPDIR="$HOME/.cache/cargo-mutants"', the fn-hash piped into 'cargo mutants --file {}', 'cargo llvm-cov nextest --workspace --json'), the exact CRAP formula, threshold 30, and concrete artifact formats like the REQ-ID comment placement. | 5 / 5 |
Workflow Clarity | Layers are explicitly sequenced cheapest-first ('do first', 'After implementation is green', 'Run it after the mutation pass') with strong validation and recovery loops: the false-green 'unviable' mutants check against mutants.out/log with a re-run remedy, and failing properties route through minimize-then-fix-or-STOP with exhaustive legal/illegal triage outcomes. | 5 / 5 |
Progressive Disclosure | Section headers (Layer 0–4, smells) navigate well, but ~250 lines of Layer 3/4 implementation detail (CRAP join rules, mutation blind spots) are inlined where reference files could carry them, and the cited scripts/crap-report.bb, scripts/complexity-gate.bb, and scripts/quality-gate.bb do not exist in the bundle. | 3 / 5 |
Total | 17 / 20 Passed |