Content
92%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a dense, operational workflow with executable commands, explicit validation gates, and a copy-paste report template — strong on actionability and workflow clarity. Its weakness is progressive disclosure: nearly everything is inline in one long file and the one external reference (api-conventions.md) is not bundled.
Suggestions
Move the Breaker and API Analyst persona definitions (the mindset bullet lists) into a referenced file such as references/personas.md and link to it, keeping SKILL.md a lean overview.
Verify and bundle the referenced api-conventions.md (e.g., under references/) — the path '.claude/skills/review/api-conventions.md' points outside the skill directory and the file is absent, which would break the API Analyst sub-agent prompt at runtime.
Extract the full report template and the severity/classification tables into a referenced file (e.g., references/report-format.md) to reduce inline bulk and improve navigation.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Quotes operational content like 'git fetch origin main && git log origin/main..HEAD --oneline && git diff --stat origin/main...HEAD' and '[SEVERITY] file:line — description — suggested fix'; it assumes Claude's competence (no git/distributed-systems primers) and every section is actionable rather than padded, matching 'Lean and efficient; every token earns its place'; not 2 because it avoids explanatory filler Claude already knows. | 3 / 3 |
Actionability | Quotes concrete executable commands ('git diff origin/main...HEAD -- <file1> <file2> ...', 'git diff --name-only origin/main...HEAD | grep -E \.api\.md$') and a full copy-paste report template; guidance is fully executable and specific, matching 'Fully executable code/commands; copy-paste ready'. | 3 / 3 |
Workflow Clarity | Quotes a 9-step sequence with explicit checkpoints — 'Empty diff gate: Zero changed files -> report ... and stop', 'Size gate: >10,000 lines changed -> ask user to narrow scope', and the 'High-confidence gate' four-point checklist; feedback loops (dedup/classify, re-validate, evidence gate) are present, matching 'Clear sequence with explicit validation steps'. | 3 / 3 |
Progressive Disclosure | Quotes 'cat .claude/skills/review/api-conventions.md' as a one-level-deep, clearly signaled reference, but the skill is otherwise a monolithic single file with personas, the full report template, and severity tables inline, and the referenced api-conventions.md is not present in the bundle; matches 'Some structure but could be better organized; content that should be separate is inline'; not 3 because split-worthy content stays inline and the reference target is missing. | 2 / 3 |
Total | 11 / 12 Passed |