Content
87%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 tight, highly actionable body: concrete commands, exact prerequisites, and unusually good non-obvious detail (config-file workaround, reporter shape normalization). The one real gap is the absence of result-validation and error-recovery steps for what is a batch operation, which caps workflow clarity.
Suggestions
Add a post-run validation step in Output handling, e.g. "Check route_count > 0 and that ci-result.json parsed; if 0 routes or a missing file, re-run with a lower --max-routes or higher --page-timeout before reporting."
Include a brief failure path for the subprocess timeout (what to do when the 600s --timeout guard trips mid-audit, e.g., reduce route cap or raise the timeout).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence — it never explains what Lighthouse or CWV is, and every line carries non-obvious information (the 25k QPD quota figure, "unlighthouse-ci has no --max-routes flag", the jsonSimple vs jsonExpanded reporter shapes). This matches the 'lean and efficient; every token earns its place' anchor; nothing reads as padding. | 5 / 5 |
Actionability | The routing table gives copy-paste-ready commands with concrete flag examples ("/seo unlighthouse <url> --device desktop", "--max-routes 50 --output-dir ./reports"), prerequisites are exact (install.sh, Node 18+), and outputs are specified field-by-field (route_count, aggregate_scores, routes). This matches the 'fully executable; copy-paste ready commands covering common cases' anchor. | 5 / 5 |
Workflow Clarity | The sequence is present (Prerequisites → Routing → Output handling) and some validation is noted as delegated to the wrapper ("url_safety pre-flight and subprocess timeout management"), but this is a batch operation (up to 200 routes) with no explicit post-run validation of results or error-recovery guidance — e.g., nothing on what to do if route_count is 0, ci-result.json is missing, or the audit times out. Per the batch-operation guideline, missing validation/feedback loops caps workflow clarity at 3; it is above the score-2 anchor because the sequence and wrapper-level pre-flight checks are clearly stated. | 3 / 5 |
Progressive Disclosure | The body is under 50 lines with no external bundle files, and the guidance says such skills can score 5 with well-organized sections alone. Sections (Prerequisites, Routing, Output handling, Cross-skill delegation) are clearly headed, and the delegation pointers to other skills (seo-google psi/crux, lcp_subparts.py) are clearly signaled one-level references, matching the 'clear overview, well-signaled' anchor. | 5 / 5 |
Total | 18 / 20 Passed |