Content
90%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.
A dense, high-quality design-system reference: terse tables, executable commands, a complete page template, explicit screenshot verification, and a checklist. The main weaknesses are a workflow that is implied across sections rather than sequenced, and a bundle-inconsistent reference to scripts/ai-context.py that does not exist alongside SKILL.md.
Suggestions
Fix the dangling reference: the body instructs 'run python3 scripts/ai-context.py frontend' as the first step, but no scripts/ directory exists in the skill bundle — either include the script or remove/repoint the instruction.
Consolidate the operating workflow into one explicitly numbered sequence near the top (run ai-context → apply tokens/ss-* classes and components → verify with four-look screenshots → run the checklist) instead of distributing it across the intro, 'Verifying a change', and the closing checklist.
Consider moving the legacy-alias migration table and the full ss-* class catalog into references/STYLE_GUIDE.md or a second reference file so SKILL.md stays a lean overview, consistent with its own 'names, does not copy' principle.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and reference-grade — token tables, class catalogs, an anti-pattern table — with no padding and no explanation of concepts Claude already knows; it even states 'This file names tokens and classes and says when to use them. It does not copy their values.' Every token earns its place, matching the lean-and-efficient anchor rather than the merely-efficient level 4. | 5 / 5 |
Actionability | Guidance is copy-paste ready: an executable grep command to list ss-* classes, a complete TSX page template, a component API table with exact props and variants, and concrete verification commands ('?theme=clean', 'run the dev server... with the ui command'). It exceeds the 'mostly executable' level 4 anchor by covering the common cases with fully executable examples. | 5 / 5 |
Workflow Clarity | The workflow is present and validated — 'Before acting, run python3 scripts/ai-context.py frontend', 'Screenshot the page in all four looks... Look at the screenshots before reporting done', and a closing checklist — but it is distributed across sections rather than one explicitly sequenced flow with error-recovery loops. This fits 'clear sequence with most checkpoints present; minor validation gaps', not the level 5 anchor's coherent validate-fix-retry structure. | 4 / 5 |
Progressive Disclosure | Structure is good: the reasoning/reference material is split one level deep into references/STYLE_GUIDE.md (which exists in the bundle), clearly signaled as 'Design rules and the reasoning behind them'. However the body's primary entry step references scripts/ai-context.py, which is absent from the bundle (no scripts/ directory), a dangling reference that keeps it below the level 5 'well-signaled, easy navigation' anchor. | 4 / 5 |
Total | 18 / 20 Passed |