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 body is information-dense with genuinely valuable post-training knowledge (2025-2026 Google rich-result retirements, @graph validation nuances, adult-product requirements) and concrete templates, but it is weakened by dead bundle references, inlined content that duplicates its own reference files, and boilerplate JSON-LD templates Claude could generate unaided.
Suggestions
Remove or externalize the three inline JSON-LD templates (Organization, LocalBusiness, Article) — they restate standard schema structures Claude already knows and duplicate the referenced schema/templates.json.
Fix the dead references: '../seo/references/schema-types.md' and 'schema/templates.json' do not exist in the bundle — either ship those files or replace the pointers with inline content, and trim the inline ACTIVE list to a few exemplars plus the pointer.
Make "Validate output before presenting" concrete: specify the validation checks or command to run and the fix-and-retry loop when validation fails.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The core is dense, non-redundant post-training knowledge (2025-2026 retirement dates, @graph inheritance rules) that Claude cannot know, and most time-sensitive dates sit in the deprecation/status sections. However, ~70 lines of standard JSON-LD templates (Organization, LocalBusiness, Article) restate schema structures Claude already knows, and the full ACTIVE list re-inlines what "Read ../seo/references/schema-types.md for the full list" points to — more than minor trimming, so anchor 3 rather than 4. | 3 / 5 |
Actionability | Concrete and mostly executable: a specific detection command ("${CLAUDE_PLUGIN_ROOT}/scripts/claude-seo run parse_html.py --url <url> --json"), copy-paste-ready templates with explicitly justified placeholders, a scenario→action error table, and named output files (SCHEMA-REPORT.md, generated-schema.json). Not 5 because the primary parse command depends on a script absent from the bundle, and "Validate output before presenting" names no mechanism. | 4 / 5 |
Workflow Clarity | Detection, Validation, and Generation are clearly sequenced, with a validation checkpoint ("Validate output before presenting") and an Error Handling table mapping failures to recovery actions (e.g., invalid JSON-LD → parse, report, provide corrected output). Not 5 because the generation-time validation step lacks an explicit method or fix-and-retry loop — it says to validate but not how or what to do on failure. | 4 / 5 |
Progressive Disclosure | Sections are well organized and references/deprecated-types-2024-2026.md is a valid, clearly signaled one-level-deep pointer, but two other referenced paths — "../seo/references/schema-types.md" and "schema/templates.json" — do not exist in the bundle, and the inline ACTIVE list and templates duplicate what those files should hold. Anchor 3: structure is present but navigation is undermined by dead references and content that belongs in the referenced files. | 3 / 5 |
Total | 14 / 20 Passed |