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-organized router/index skill body with concrete procedures, checklists, and a strong output format, and no token waste on concepts Claude already knows. Its two real defects are the broken disclosure layer — all 20+ referenced rule files and AGENTS.md are missing from the bundle — and local flaws like the dangling 'For Formal Reviews' intro and duplicated rule listings between the priority table and Quick Reference.
Suggestions
Ship the `rules/` directory (all 20+ referenced rule files) and `AGENTS.md` in the bundle, or inline the rules' content — as delivered, every 'Read these rule files' instruction fails because the files do not exist.
Fix the 'For Formal Reviews' section, which ends with a colon ('When performing a formal review of schemas, queries, or data ingestion:') and then drops into unrelated content; either complete its list or remove the section.
Collapse the duplication between 'Rule Categories by Priority' and 'Quick Reference' — both enumerate the same 28 rules with one-line summaries — into a single table (e.g. add the one-line description and impact to the priority table).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is efficient and assumes Claude's competence — it never explains what ClickHouse or columnar storage is and gets straight to procedures — but the 'Rule Categories by Priority' table and the 'Quick Reference' section both enumerate the same 28 rules with one-line summaries, a clear duplication that could be trimmed. Not anchor 5 because of that redundancy and the low-value 'Rule File Structure' section. | 4 / 5 |
Actionability | Guidance is highly concrete — ordered rule-file reading lists per review type, specific checklists, a copy-paste output-format template, and a citation pattern ('Per `rule-name`...') — which would merit anchor 5. However, every referenced path (20+ `rules/*.md` files and `AGENTS.md`) is absent from the bundle, so the primary instruction 'read these rule files' is not actually executable as shipped. | 4 / 5 |
Workflow Clarity | A clear numbered priority sequence (check rules → apply and cite → fall back to knowledge → web search → always cite source) plus per-review-type procedures and an output format whose 'Rules Checked / Compliant / Violation' structure functions as a verification report. Docked from 5 for the dangling 'For Formal Reviews' section that ends in a colon with no following list, leaving that workflow entry point incoherent. | 4 / 5 |
Progressive Disclosure | As written, the design is good — one-level-deep references to `rules/` files, clearly signaled and grouped by review type — but scored against the actual bundle structure, no `rules/` directory or `AGENTS.md` exists at all, so every reference is broken and the detailed content the SKILL.md is supposed to disclose is missing. The structure is better organized than anchor 3's example, but the broken references prevent the anchor 4-5 navigation experience. | 3 / 5 |
Total | 15 / 20 Passed |