Content
92%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.
An exemplary instruction-only skill body: lean, phase-structured with explicit gates and checklists, and clean progressive disclosure into three well-scoped reference files that all exist. The only gap is the absence of a small inline worked example to ground the description-writing guidance.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every line is a directive or checkpoint; it never explains concepts Claude already knows (no 'what is MCP' padding) and uses tables and terse bullets so every token earns its place. | 5 / 5 |
Actionability | Concrete, executable directives throughout ('Name tools `verb_noun`. Keep the full name at or under 64 characters', 'Return recoverable errors with `isError: true` and a next step', 'Use **stdio** for local desktop tools'), but the body includes no worked example (e.g., a sample what/returns/when tool description) to make the guidance copy-paste concrete. | 4 / 5 |
Workflow Clarity | The Design → Build → Harden → Test phases are clearly sequenced with per-phase 'Done when:' criteria, explicit 'Phase gates' validation checkpoints between phases, and a closing quality checklist — matching the feedback-loop/checklist pattern the rubric rewards. | 5 / 5 |
Progressive Disclosure | The body is a lean overview that clearly signals three one-level-deep references ('For naming, pointer wording, schemas, URIs, and anti-patterns, load `references/tool-design.md`' and parallels), each naming its contents and when to load it; all three referenced files exist in the bundle. | 5 / 5 |
Total | 19 / 20 Passed |