Content
100%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 lookup skill body: lean token-efficient tables with links, unambiguous recommendation procedure, explicit decision splits for ambiguous pairs, and concrete mismatch heuristics for catching hand-rolled alternatives. Nothing is padded, nothing needs external files, and every section is immediately actionable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is pure lookup tables and decision guidance — it explains nothing Claude already knows (no library tutorials, no concept explanations), and every prose note (e.g. "A simple hover or fade doesn't need it — plain CSS transitions are the right tool there") is decision-relevant, matching anchor 5's "every token earns its place". | 5 / 5 |
Actionability | Concrete, executable guidance throughout: "Look at `package.json` first", "Recommend one library, state what it's for in one sentence", a named library with a link for every task, explicit split rules (Liveline vs recharts, clsx vs cva), and specific mismatch heuristics — for an instruction-only skill this fully covers the common cases (anchor 5). | 5 / 5 |
Workflow Clarity | The simple-skill exception applies: the four-step procedure (identify task → check what's installed → recommend one → explicit fallback when uncovered) is unambiguous, includes a dependency-churn guardrail ("don't churn the dependency without being asked"), and involves no destructive or batch operations that would require validation checkpoints. | 5 / 5 |
Progressive Disclosure | No bundle files exist or are needed — the ~80-line body is dense tabular lookup data that constitutes the skill's entire value, appropriately kept inline and organized into clear, navigable sections (UI components, motion, charts, interaction, state/styling, mismatches) with no buried references. | 5 / 5 |
Total | 20 / 20 Passed |