Content
80%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 command reference: every command is executable with exact flags, defaults, and error-recovery actions, and detail is correctly pushed to a complete, well-signaled set of reference files. The main gaps are missing validation checkpoints before batch operations (capped by the rubric) and dated time-sensitive notes scattered outside the Technical Notes section.
Suggestions
Add an explicit validation step before batch operations: for index-batch/inspect-batch, verify the URL file (one URL per line, count against the 200/2,000 limits) and confirm the submission scope with the user before executing.
Consolidate time-sensitive dated information (2026 GSC AI-surfaces notes, GA4 AI Assistants channel launch dates) into a dedicated 'Data caveats by date' or deprecated-changes section alongside the existing Technical Notes, so stale dates are easier to audit.
Break the GSC 'Which flag' guidance into a short bulleted rule list (--property vs --domain vs neither) to replace the current dense run-on paragraph.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is an efficient command reference with essentially no explanation of concepts Claude already knows — each command gets one script line, defaults, and quota. It is not a 5 because several time-sensitive dated items sit outside the deprecation-style Technical Notes section (the 2026 AI-surfaces blockquotes, the GA4 'AI Assistants' ~2026-05-13 note) and a few passages like the GSC 'Which flag' paragraph are dense run-ons that could be tightened. | 4 / 5 |
Actionability | Every command ships a copy-paste executable invocation with exact flags (e.g. '"${CLAUDE_PLUGIN_ROOT}/scripts/claude-seo" run gsc_query.py --domain <audited-domain> --json'), stated defaults, quotas, a concrete error-scenario/action table, and a fully specified 3-step report workflow. This matches the anchor for fully executable, copy-paste-ready commands covering the common cases. | 5 / 5 |
Workflow Clarity | Sequencing is generally clear — credentials are checked up front with a concrete command, the detected tier is communicated before running, and the error table maps failures to recovery actions. However, the batch operations (inspect-batch at 2,000/day and index-batch submitting up to 200 URLs to Google's Indexing API) have no pre-execution validation or verification steps — no file-content check, URL-count check, or confirm-before-submit loop — which the rubric explicitly caps at 3 for batch/destructive workflows. | 3 / 5 |
Progressive Disclosure | The SKILL.md is a well-organized overview: a quick-reference command table up top, per-family sections, and per-command '**Reference:**' labels pointing to one-level-deep files. All 12 referenced files in references/ and all 3 templates in assets/templates/ exist on disk, and the reference files contain no nested references, matching the anchor for a clear overview with well-signaled one-level-deep references. | 5 / 5 |
Total | 17 / 20 Passed |