Content
71%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 well-engineered operational skill: the research workflow is concrete and executable with specific tools, paths, and extraction checklists, and the writing never over-explains. Its main weakness is duplication — the full profile template and per-tool details are inlined in SKILL.md despite dedicated reference files existing for exactly that content.
Suggestions
Move the full profile template (the ~140-line block under "Each profile follows this structure") into references/templates.md, keeping only a section outline in the body — the existing "For the full profile and summary templates" pointer already promises this split.
Trim the Phase 2 subsections to tool name + one-line purpose, since tool-reference.md already documents each tool's purpose, when-to-use, and key output; keep the save-path and cross-referencing instructions inline.
Add an explicit validation checkpoint in Phase 1 (e.g., verify each scrape returned usable content and re-fetch or fall back to WebFetch on failure) rather than deferring all error handling to the reference.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The prose itself is lean — no basic-concept explanations, efficient tables — but the ~140-line full profile template is inlined in the body even though a dedicated templates.md reference exists, and the Phase 2 per-tool listings overlap tool-reference.md's per-tool documentation. Not 4 because a third of the body could be trimmed or moved; not 2 because there is no padding or over-explanation of known concepts. | 3 / 5 |
Actionability | Fully actionable for the common cases: exact MCP tool names per step ("firecrawl_map → competitor URL", "dataforseo_labs_google_ranked_keywords"), exact save paths with naming conventions ("competitor-profiles/raw/<competitor-slug>/<YYYY-MM-DD>/scrapes/<page-name>.md"), per-page extraction tables, and a default-depth heuristic ("Default to quick scan unless... 3 or fewer"). As an instruction-only skill the guidance is concrete throughout, with literal example calls correctly one reference away. | 5 / 5 |
Workflow Clarity | The sequence is clear and well-phased (assessment → stack selection → scrape → SEO → synthesis → output → summary) with real checkpoints (persist raw data before extracting, "never estimate" unavailable SEO metrics, cross-reference claims in Phase 3, check active MCPs before starting). Not 5 because error-recovery loops are delegated to the reference rather than stated as explicit validate-retry steps in the body. | 4 / 5 |
Progressive Disclosure | References are real, clearly signaled, and one level deep, but the body inlines the full profile template right after pointing to references/templates.md ("For the full profile and summary templates: See references/templates.md"), and the Phase 2 tool-by-tool listings duplicate tool-reference.md's structure. Not 4 because content that clearly belongs in the existing reference files occupies roughly a third of the body. | 3 / 5 |
Total | 15 / 20 Passed |