Content
48%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 has a strong, clearly sequenced workflow with real verification checkpoints and some concrete search tooling, but it is significantly overlong and padded with placeholder response templates and repeated content. Empty comment-only code blocks and the absence of any progressive-disclosure structure (no reference files despite 450 lines) are the main weaknesses.
Suggestions
Cut the placeholder response templates and duplicated sections ('When to Use', 'Quick Reference', 'The Bottom Line') — Claude does not need scripted replies; keep the issue-category and root-cause tables plus the tool patterns, targeting roughly half the current length.
Replace comment-only bash blocks with actual commands or drop them, and keep the concrete Glob/Grep patterns as the primary actionability payload.
Split the Common Patterns, verification checklists, and Best Practices into reference files under references/ (e.g., references/patterns.md, references/verification.md), leaving SKILL.md as a lean overview that links to them one level deep.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~450-line body is noticeably padded: placeholder response templates ("I can see the screenshot showing [describe what you observe]"), a 'When to Use' section that restates the frontmatter trigger block, a 'Quick Reference' table repeating the 'Common Patterns' section, and a 'The Bottom Line' restating the core principle. Redundancy is pervasive rather than occasional, matching anchor 2. | 2 / 5 |
Actionability | There is genuinely concrete guidance (Glob patterns like "**/*settings*.{tsx,jsx,ts,js,vue,svelte}", Grep regexes, `npm run dev`), but several bash blocks contain only comments with no command ("# Search for component files related to the issue") and most markdown blocks are placeholder response templates rather than executable instruction — matching the 'some concrete guidance but incomplete / pseudocode' anchor. | 3 / 5 |
Workflow Clarity | A clear five-phase sequence (analyze → investigate → root-cause → implement → verify) with a verification checklist, responsive/state checks, comparison against the original screenshot, and an explicit user-confirmation loop. It falls short of anchor 5 because validation is a manual checklist rather than an automated validate-fix-retry loop, though the batch 'fix everywhere' flow does include the verification phase. | 4 / 5 |
Progressive Disclosure | No bundle files exist or are referenced; everything — pattern libraries, checklists, best-practice examples — is inlined in a single 450-line file. Internal section headers are good (above anchor 2's headerless wall), but content that belongs in separate reference files is inline with no reference split, matching anchor 3. | 3 / 5 |
Total | 12 / 20 Passed |