Content
75%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, highly actionable audit skill: concrete grep-based checks, clear severity levels, an explicit verification checklist, and a defined output format. The main weaknesses are minor — an unbound $TARGET variable, an unspecified 0-100 scoring methodology, some rule/checklist redundancy, and contrast verification lacking a concrete command.
Suggestions
Bind the target variable explicitly (e.g., note that $TARGET in the grep commands is the resolved $ARGUMENTS path) so every command is executable as written.
Define how the per-category 0-100 scores and the overall verdict thresholds (PASS / PASS WITH NOTES / NEEDS WORK / FAIL) are computed, so Step 6 is reproducible.
Add a concrete verification method for the 4.5:1 contrast criterion (e.g., a named tool or calculation step), since it is the one accessibility check with no executable command.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational — checklist tables, executable grep commands, and a severity taxonomy — with no explanation of concepts Claude already knows (no 'what is WCAG/Tailwind' padding). Minor trimmable redundancy keeps it below anchor 5: the Rules section repeats Verification Protocol items (file:line evidence, token-first checking), and the Anti-Rationalization table is extra framing beyond the core audit procedure. | 4 / 5 |
Actionability | Concrete, executable grep commands cover the common cases (hardcoded colors, missing alt text, unlabelled inputs, fixed widths, missing responsive images) and the output template is copy-paste ready. Minor gaps hold it at anchor 4 rather than 5: '$TARGET' is used in commands but never bound to '$ARGUMENTS', 'Calculate scores per category (0-100)' specifies no scoring method, and the 4.5:1 contrast check lists no verification command or tool. | 4 / 5 |
Workflow Clarity | A clearly sequenced 6-step workflow with an explicit Verification Protocol checklist ('Before claiming the review is complete:') and a false-positive re-check (Rule 7), which fits anchor 4's 'clear sequence with most checkpoints present'. It falls short of anchor 5 because the per-category scoring step is undefined and there are no error-recovery/feedback loops for ambiguous findings. | 4 / 5 |
Progressive Disclosure | No bundle files exist, and the ~200-line body is well-sectioned with headers, tables, and code blocks so all inline content is scannable and appropriately placed — matching anchor 4's 'good structure; most content is appropriately placed'. It does not reach 5 because it exceeds the under-50-line simple-skill case, and the accessibility/UX checklists could arguably be split into one-level-deep reference files to keep SKILL.md leaner. | 4 / 5 |
Total | 16 / 20 Passed |