Content
83%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 exemplar of lean, highly actionable instruction-only content: exact tokens, components, variants, and anti-pattern match patterns with no wasted tokens. Its main gap is workflow-level — the fix-application path (batch edits, default on) lacks any post-fix validation step — and its 74-line inline token catalog leaves modest room for reference-file splitting.
Suggestions
Add a post-fix validation step to the workflow, e.g. "5. After applying fixes, run the typecheck/build and re-scan the diff to confirm no violations remain".
Consider moving the full token catalog (text/surface/border/z-index/badge families) to a references/tokens.md file and keeping only the most common tokens inline, bringing SKILL.md under the simple-skill threshold.
Clarify in the Steps how findings should be reported when fix=false (e.g., grouped by file or by rule) so the propose-only path is as well-specified as the apply path.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a lean, checklist-style rule catalog with zero padding and no explanation of concepts Claude already knows — every line states a rule, a token, or an anti-pattern. It matches anchor 5 (lean and efficient, assumes Claude's competence); anchor 4 would require identifiable over-explanation to trim, and none is present. | 5 / 5 |
Actionability | Guidance is fully concrete: exact file paths to read ("packages/emcn/src/index.ts", "apps/sim/app/_styles/globals.css"), exact token names with values ("--z-dropdown (100)"), exact component/variant/prop names, and concrete match patterns to flag ("text-gray-*", "#hex", "rgb()", "h-[Npx] w-[Npx]"). Per the rubric's scoring notes, an instruction-only skill with this level of specific, actionable guidance scores at anchor 5. | 5 / 5 |
Workflow Clarity | The 4-step sequence (read barrel, read globals.css, analyze scope against every rule, apply or propose fixes) is clearly ordered, but the workflow ends at "apply the fixes" with no validation checkpoint (e.g., typecheck, build, or visual re-verification). Since fix=true is the default and scope can be "whole codebase" (a batch edit operation), the missing validation step caps workflow clarity at 3 per the judging guidelines. | 3 / 5 |
Progressive Disclosure | The single-file body (~74 lines) is well organized with clear section headers and appropriately signals external codebase paths inline. However, it exceeds the under-50-line simple-skill threshold, and content such as the full token catalog and the dense Mobile rules paragraph could be split into a reference file. This matches anchor 4 (good structure, most content appropriately placed, minor organization gaps) rather than 5. | 4 / 5 |
Total | 17 / 20 Passed |