Content
71%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 well-structured reference skill: the body stays an overview, defers mechanics to sibling skills, routes all detail to 8 real, one-level-deep reference files, and enforces its guidance with concrete BAD/GOOD anti-patterns. Its main costs are duplicated trigger sections (When to Apply / Use When / When NOT to Use overlap) and the lack of any inline example for key remapping or custom mapped types, plus a validation command not yet tied into a sequence.
Suggestions
Merge the 'When to Apply' bullet list and 'Use When' quote list into one section (the bullets and quotes restate the same triggers), and fold 'When NOT to Use' into 'Out of Scope' instead of repeating the deep-readonly handoff twice.
Add one short inline example each for key remapping (`as` clause) and a custom mapped type, since these are the two tasks where users cannot yet judge from the body alone what the skill will produce.
Turn 'Quick Commands' into a step: 'after applying a transform, verify with `npx tsc --noEmit`' so the type-check acts as an explicit validation checkpoint in the workflow.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body avoids explaining concepts Claude already knows (it defers mapped-type mechanics to a sibling skill) and its code examples are tight, but the triggers are restated three times — "When to Apply" (8 bullets), "Use When" (4 quotes), plus the description — and "Out of Scope" repeats verbatim in "When NOT to Use" (deep-readonly named twice). This is the 3 anchor ("mostly efficient but includes some unnecessary explanation or could be tightened") rather than 2, since nothing is conceptually padded — the redundancy is structural duplication that could be merged. | 3 / 5 |
Actionability | The three anti-patterns give executable BAD/GOOD TypeScript pairs (e.g., `type UserUpdate = Partial<User>` vs a hand-written interface), and the References table maps each need to a concrete file, but the body itself gives no inline example of the two tasks users most need shown — key remapping and custom mapped types are only pointed at. This fits the 4 anchor ("mostly executable guidance; concrete code or commands with minor gaps") rather than 5, since the common cases rely on navigation rather than inline copy-paste coverage. | 4 / 5 |
Workflow Clarity | The decision path is clear: Mindset establishes built-in-first, combine-don't-invent rules; When to Apply lists triggers; and the References table routes each need to its file, with `npx tsc --noEmit` offered as verification. It sits at the 4 anchor rather than 5 because the validation command is presented as a loose "Quick Commands" item rather than an integrated step (e.g., "apply the transform, then type-check"), leaving a minor checkpoint gap; no destructive/batch cap applies since this is a non-destructive selection skill. | 4 / 5 |
Progressive Disclosure | The body is a genuine overview: mindset, triggers, scope boundaries, anti-patterns, and a References table routing to all 8 bundle files — each of which exists in `references/` exactly as listed, is one level deep, and is labeled with what it covers. This matches the 5 anchor (clear overview with well-signaled one-level-deep references, content appropriately split, easy navigation). | 5 / 5 |
Total | 16 / 20 Passed |