Content
65%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is token-efficient and well structured with useful red-flag self-checks, but it leans on principles over concrete worked examples and leaves half of its reference bundle unlinked. Tightening actionability and surfacing all bundle files would lift the weakest dimensions.
Suggestions
Add one short worked example showing a real finding rendered in the strict output format (severity, file, why, fix) to make guidance copy-paste ready.
Link the remaining bundle files (lenses.md, report.md, request-template.md) from a References section so all reference material is discoverable one level deep.
Replace the high-level checklist with a short ordered review workflow (e.g. skim diff -> classify findings -> evidence each -> output grouped by severity) with an explicit verification step.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a lean, bullet-driven overview that assumes Claude's competence and avoids explaining what code review is; every section earns its place despite a few terse fragments. | 3 / 3 |
Actionability | A concrete output-format template and severity taxonomy are provided, but most guidance is principle-style ('Find bugs, flaws, design errors') with no worked example, landing at 'some concrete guidance but incomplete'. | 2 / 3 |
Workflow Clarity | A mandatory checklist and 'Stop if' red-flag checkpoints give some sequence and self-correction, but there is no clear end-to-end review workflow with explicit validate/decide checkpoints. | 2 / 3 |
Progressive Disclosure | Sections are well organized and two references are clearly signaled one level deep, but lenses.md, report.md, and request-template.md exist in the bundle yet are not surfaced from the body. | 2 / 3 |
Total | 9 / 12 Passed |