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.
A thorough, well-sequenced audit workflow with strong concrete artifacts (scoring rubric, output template, generation rules), undermined by a monolithic structure that inlines schema.org boilerplate Claude already knows and references a fetch script that is missing from the bundle. Tightening it means moving per-type property lists to reference files and fixing or removing the fetch_page.py dependency.
Suggestions
Move the per-type property lists (Step 3) into references/ files (e.g., references/organization.md, references/article.md, references/product.md) and keep one-line summaries with links in SKILL.md, cutting the body roughly in half.
Either bundle scripts/fetch_page.py or replace the instruction with an inline executable approach (e.g., a curl + python json-ld extraction snippet) so the detection step is not a broken reference.
Add an explicit validate → fix → re-validate loop after Step 2 (e.g., 'if a schema fails validation, regenerate it with the missing properties before scoring') to close the workflow's feedback gap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Steps 3 and 6 spend ~180 lines enumerating standard schema.org required/recommended properties (Article needs headline/datePublished/author, PostalAddress fields, a full Organization JSON template, a WebSite+SearchAction block) that Claude already knows, matching 'mostly efficient but includes some unnecessary explanation or could be tightened'. Not 2 because the GEO-specific annotations (knowsAbout as a strong signal, sameAs priority order, Dec 2025 JS-injection guidance, deprecation table) are genuinely non-obvious and earn their tokens; not 4 because the boilerplate property lists and templates are a large share of the document. | 3 / 5 |
Actionability | Concrete commands ("python3 ~/.claude/skills/geo/scripts/fetch_page.py <url> page"), a copy-paste Organization JSON-LD template, a 0-100 scoring rubric with per-criterion point rules, and a full output-format skeleton make the guidance mostly executable. Not 5 because the only command is a script that is not in the bundle (no scripts/ directory exists), and generation guidance for types other than Organization is property-list-only with no complete examples. | 4 / 5 |
Workflow Clarity | A clear 6-step sequence (detect → validate → identify gaps → generate → report) with a numbered "How to Use" summary and an 8-point validation checklist in Step 2; the scoring rubric and output template define explicit completion criteria. Not 5 because there is no explicit feedback loop (e.g., 'if validation fails, fix the JSON-LD and re-check before generating the report') — checkpoints are listed but recovery steps are implicit. | 4 / 5 |
Progressive Disclosure | Section structure is clear and well-ordered, but everything is inlined in a single ~370-line file: the per-type property references in Step 3 are exactly the content that belongs in separate reference files, and the sole external reference (fetch_page.py) points to a path that does not exist in the bundle (no scripts/ directory), so it doesn't resolve. This matches 'some structure but... content that should be separate is inline' rather than 2, since headers and navigation are good, and not 4 because the one bundle reference is broken and no reference files exist at all. | 3 / 5 |
Total | 14 / 20 Passed |