Content
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a highly actionable and well-structured review skill with clear workflows, explicit severity criteria, and a concrete output template. Its main weakness is length — the seven passes contain significant detail that could be split into referenced files for better progressive disclosure, and there's some redundancy between the 'always Critical' preamble and the individual pass descriptions. Despite the verbosity concern, nearly every line serves a purpose in preventing false positives/negatives during code review.
Suggestions
Extract the seven detailed pass definitions into a separate reference file (e.g., PASSES.md) and keep only a summary checklist in SKILL.md to improve progressive disclosure and reduce the monolithic feel.
Consolidate the 'always Critical' list with the per-pass details to eliminate redundancy — either keep only the per-pass version or make the preamble a terse summary table that references the passes.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is thorough and most content earns its place — the seven passes, severity definitions, and report template are all necessary. However, there's some redundancy: the 'always Critical' list in the preamble repeats nearly verbatim in Passes 2-5, and some explanatory framing (e.g., the 'When to use' section's cross-references to other skills) could be tighter. Overall mostly efficient but not maximally lean. | 2 / 3 |
Actionability | Extremely actionable: each pass has concrete, specific criteria with exact file paths, naming conventions, tool names, and MCP function signatures. The report template is copy-paste ready with a fixed contract. Verification commands are fully specified (`pnpm lint && pnpm db:check:migrations && ...`). The severity rules are unambiguous with concrete examples (e.g., 'new Drizzle*Repository(...) outside the composition root'). | 3 / 3 |
Workflow Clarity | The workflow is clearly sequenced: identify inputs → load context (4 ordered steps) → run seven named passes → verify evidence → produce report → post/return based on mode. Validation checkpoints are explicit (confirm PR body shows green suite, run checks if no evidence). The verdict logic includes clear decision branches (Ready/Needs changes/Blocked) with specific follow-up actions for each. Error recovery is addressed (e.g., ambiguous input → ask, plugin not loaded → note and continue). | 3 / 3 |
Progressive Disclosure | The skill is a single monolithic file with no bundle files to offload detail into. At ~300+ lines, the seven detailed passes with their sub-bullets could benefit from being split into referenced files (e.g., a checklist file per pass, or a separate conventions reference). The content is well-organized with clear headers, but the sheer volume in one file works against progressive disclosure principles. | 2 / 3 |
Total | 10 / 12 Passed |