Content
96%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 skill: token-efficient, fully actionable with concrete IR examples and generated-code mappings, and a clearly sequenced workflow with real validation commands and error-recovery feedback. The only structural nit is that the dense field/projection reference material is all inline with no external split.
Suggestions
Consider moving the full class-field and projection-step reference tables into a one-level-deep reference file (e.g., references/ir-fields.md), keeping SKILL.md to the workflow, decision framework, and a few key examples.
Optionally add a short 'Common patterns' cross-link to worked domain-map.json examples for new tools, so edge cases (acknowledgeSingle, find) don't need to be fully specified inline.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense tables and examples with essentially no filler: no explanation of concepts Claude already knows, no library-choices digressions, no padded prose. Each section (arg routing, projections, cache) delivers rules and examples only, matching the lean anchor. | 5 / 5 |
Actionability | Guidance is executable throughout: concrete JSON snippets for classes/bindings/cache, field→example tables showing exact generated code (e.g., `name: \`projects/${this.projectId}/screens/${screenId}\``), a decision framework keyed on real tool properties, and copy-paste validation commands plus a real diagnostic error example. Not below 5 on any identifiable gap. | 5 / 5 |
Workflow Clarity | The pipeline is clearly sequenced (Inputs → Output → Designing Classes → Designing Bindings → Decision Framework → Validation) with explicit validation checkpoints: five runnable commands and a sample validation error that lists available properties, giving a concrete fix-and-retry feedback loop. The Decision Framework section doubles as a checklist for new tools. | 5 / 5 |
Progressive Disclosure | Sections are well-organized with tables, callouts, and clear navigation; there are no nested or buried references and no orphaned bundle paths. Not a 5 because ~190 lines of dense IR reference (class-field table, arg routing, projection semantics) are all inlined in SKILL.md; the projection/arg-routing tables could plausibly live in a one-level-deep reference file to keep the overview leaner. | 4 / 5 |
Total | 19 / 20 Passed |