Content
86%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-organized, highly actionable advisory skill body with strong mode dispatch and validation loops for the auto-apply mode. Its main weaknesses are redundant inline duplication of the bundle's quick-reference file and heavy reliance on `rules/` files that are absent from the evaluated bundle.
Suggestions
Ship the referenced `rules/*.md` files in the bundle (or remove/repair the ~30 dead links, e.g. `rules/procedure.md`, `rules/stacks/react/components.md`) so the 'When the code involves... | Load' navigation table actually resolves.
Replace the ~35-row inline Quick Reference table with a link to the existing `references/quick-reference.md`, keeping only the 5–8 highest-leverage rows in SKILL.md to cut duplicated context.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes Claude's competence (e.g. 'Comments rot; names get refactored with code', '0.1 + 0.2 !== 0.3') with no basic-concept padding, but the ~35-row Quick Reference table duplicates content that already ships in `references/quick-reference.md`, and the four modes are described twice (mode list plus Procedure section). | 4 / 5 |
Actionability | Guidance is highly concrete for an instruction-only skill: specific refactor targets ('One `Record<Status, { label, color, icon }>`', '`type User = z.infer<typeof UserSchema>`'), quantified thresholds (50+ lines, 4+ params, debounce ~1000 ms), exact invocations ('`Skill("code-quality", "plan")`', 'confidence ≥ 90 %'), and an explicit mode-dispatch rule from `$ARGUMENTS`. | 5 / 5 |
Workflow Clarity | The four modes are clearly sequenced with an explicit dispatch rule ('Detect the mode from the `$ARGUMENTS` first token ... default is authoring'), and simplify mode has real feedback loops ('apply one recipe at a time behind `Skill("confidence", "code") ≥ 90 %` and a scoped fast-check, reverting on failure'). The detailed 13-step authoring and 8-step review checklists are deferred to `rules/procedure.md` rather than stated inline, leaving minor validation gaps in the body itself. | 4 / 5 |
Progressive Disclosure | Signaling and one-level-deep organization are good (per-topic 'When the code involves... | Load' tables), but the body references ~30 `rules/*.md` paths (control-flow.md, stacks/react/components.md, procedure.md, etc.) that do not exist in the bundle — only `references/citations.md` and `references/quick-reference.md` are present — so navigation dead-ends. Additionally `references/quick-reference.md` is never linked from the body while its content is inlined, which is structure that should be separate. | 3 / 5 |
Total | 16 / 20 Passed |