Content
71%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 extremely lean router that concisely delegates to an `@expert-reviewer` agent, which is token-efficient but leaves the skill body thin on actionable review detail and workflow structure. The real methodology is collapsed into a single agent reference rather than surfaced as navigable reference files, weakening progressive disclosure.
Suggestions
Surface a compact outline of the 4-wave workflow (the wave names and their validation/checkpoint gates) directly in SKILL.md so the review sequence and feedback loops are visible without entering the agent.
Move the per-dimension checklists and folder hotspot mapping into dedicated files under references/ (e.g., dimensions.md, hotspots.md) and link to them from the body, so the detailed content is navigable one level deep instead of hidden inside one agent.
Add a brief "what to return / how to verify the review is complete" note so the body itself is actionable, rather than relying entirely on the agent for execution guidance.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is two lean sentences ("Invoke `@expert-reviewer`..." and a summary of what the agent contains) with no padding or explanation of concepts Claude already knows, matching anchor 5 (lean, assumes competence, every token earns its place). | 5 / 5 |
Actionability | "Invoke `@expert-reviewer`" is a concrete executable instruction, but the body provides no review specifics — all methodology, checklists, and the 4-wave workflow are offloaded to the agent, leaving key details missing; this matches anchor 3 (some concrete guidance but incomplete) rather than 4 which expects coverage of common cases. | 3 / 5 |
Workflow Clarity | The single action "Invoke `@expert-reviewer`" is unambiguous, keeping it above anchor 3, but the referenced "4-wave workflow" is hidden inside the agent with no visible sequence or validation checkpoints, so it does not reach anchor 5's explicit validation steps and feedback loops. | 4 / 5 |
Progressive Disclosure | The overview is lean and the delegation is signaled ("@expert-reviewer" is named), but the substantial content (24 dimensions of checklists, folder hotspot mapping, 4-wave workflow) is collapsed into one opaque agent with no navigable file structure, matching anchor 3 (some structure, content that should be surfaced is buried) rather than 4's well-organized split. | 3 / 5 |
Total | 15 / 20 Passed |