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 dense, highly actionable audit skill with executable commands, a concrete output contract, and a well-used reference file. The main weaknesses are length: the AI-crawler section duplicates a referenced skill's authoritative table, and several date-heavy policy passages could be extracted into references to tighten the token budget.
Suggestions
Extract the AI crawler table and the Googlebot fetch-limit / crawl-rate / AMP policy notes into a reference file (e.g., references/ai-crawlers.md, references/google-policy-notes.md) and keep only the audit-relevant summaries in SKILL.md, since the authoritative AI-crawler table is already stated to live in the seo-geo skill.
Tighten the date-stamped passages (AMP history, canonicalization two-week note, JS SEO December 2025 clarifications) to their audit decision rules, moving background detail to a reference or a clearly labeled 'recent changes' section to reduce token cost and staleness risk.
Add an explicit ordered runbook for the audit pass (e.g., 1. fetch robots.txt and run sitemap_discovery.py, 2. render check, ... 9. compile the Output report) so the sequence between the category checklists and the report template is explicit rather than inferred from section order.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The checklists are lean fragments and much of the content is genuinely current, non-obvious 2025–2026 guidance, but there is notable inline bulk that could be trimmed or deferred: the 10-row AI-crawler table inlined even though the body says "the authoritative table ... is in `seo-geo`", plus long date-stamped policy notes (AMP history, crawl-rate, JS SEO December 2025) with many specific dates that penalize conciseness per the rubric. Not 2, because there is little padding explaining concepts Claude already knows; not 4, because the AI-crawler section alone is ~50 lines of duplicated material. | 3 / 5 |
Actionability | Commands are copy-paste ready with flags and JSON output (e.g., `"${CLAUDE_PLUGIN_ROOT}/scripts/claude-seo" run sitemap_discovery.py <url> --json`, `crux_history.py <url> --form-factor PHONE --json`), the env-var allowlisting example is a complete bash block, and the output/error-handling sections give concrete report and recovery actions. This matches the fully-executable 5 anchor; the 4 anchor's 'minor gaps' would understate the coverage. | 5 / 5 |
Workflow Clarity | The nine ordered categories plus the Output section ("Score only what was measured", "a category you could not measure is reported as 'not measured', never given a number", "Show the checks behind every score") and the Error Handling table provide a clear sequence with most checkpoints present. Not 5 because there is no explicit ordered runbook of the audit steps and no validate-then-proceed feedback loop; not 3 because checkpoints are explicit rather than implicit. | 4 / 5 |
Progressive Disclosure | Good section structure with clear headers; the one bundle reference (`references/stack-cve-exposure.md`) is real and appropriately deferred to ("the current PHP critical-RCE advisory ... are in `references/stack-cve-exposure.md`"), and cross-skill pointers (seo-geo, seo-schema, seo-agentic, SECURITY.md) are one level deep. Not 5 because ~290 lines of inline detail — notably the AI-crawler table that duplicates seo-geo and the Googlebot fetch-limit/AMP policy notes — would fit better in reference files given only one reference file exists. | 4 / 5 |
Total | 16 / 20 Passed |