Content
63%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 content is a well-organized, highly actionable UX reference with concrete thresholds and GOOD/BAD code examples for nearly every law, plus useful cross-cutting dedup guidance. Its main weaknesses are token inefficiency (known-law definitions, checks repeated across three formats, a redundant Keywords line) and the absence of any progressive disclosure — the full 21-law reference lives inline in SKILL.md instead of split into reference files.
Suggestions
Split the per-law detail into one-level-deep reference files (e.g., references/cognitive-laws.md, references/visual-laws.md, references/behavioral-laws.md) and keep SKILL.md as the overview plus the Review Checklist table, with clearly signaled links to each file.
Eliminate the triple repetition of checks: pick one canonical location for the 'what to look for / review flags' material — either the checklist table or the per-law sections — and delete the other two, also removing the 'Keywords:' line that duplicates the frontmatter triggers.
Trim or drop the law 'Definition' lines (Claude already knows Miller's, Hick's, and Fitts's Laws) and keep the application-focused content: key takeaways, code examples, and review flags; add the missing BAD examples for Pareto Principle, Parkinson's Law, and Uniform Connectedness.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly dense, actionable material (examples, thresholds, flags), but it restates well-known law definitions Claude already knows ("The average person can hold approximately 7... items in working memory"), repeats each check three times (Review Checklist table, "What to look for in code", and "Review flags" sections), and pads in a "Keywords:" line that duplicates the frontmatter triggers. Not a 2 because the majority of tokens still earn their place; not a 4 because the triplication and known-concept definitions are systematic waste. | 3 / 5 |
Actionability | Concrete, executable guidance throughout: specific thresholds ("44x44px", "more than 7 top-level items", "400ms") and complete GOOD/BAD HTML and CSS examples for roughly 18 of the 21 laws. It falls short of 5 because a few laws (Pareto Principle, Parkinson's Law, Uniform Connectedness) lack complete examples, and many examples lean on project-specific tokens and companion skills (--op-space-*, optics-context, bem-structure) that may not exist in a target repo. | 4 / 5 |
Workflow Clarity | A clear sequence is given ("check the Review Checklist first to identify which laws are most relevant, then refer to the detailed sections"), both modes (review and guidance) are defined up front, and the Cross-Cutting Concerns section tells the reviewer to group related findings rather than duplicating flags. Not a 5: there is no guidance on how to prioritize or format reported findings, and the guidance-mode workflow is thin beyond 'apply UX laws when building new features'. The destructive/batch validation cap does not apply since this is a read-only advisory skill. | 4 / 5 |
Progressive Disclosure | Internal structure is good (overview, scan-friendly checklist table, three categorized law sections, cross-cutting dedup guidance), but roughly 900 lines of per-law reference material are inlined in SKILL.md with no bundle files at all — no references/, scripts/, or assets/ exist. Content that clearly belongs in separate one-level-deep reference files (e.g., per-category law references) is inline, matching the anchor-3 pattern rather than a 4 or 5. | 3 / 5 |
Total | 14 / 20 Passed |