Content
73%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 content is a sophisticated, well-sequenced workflow with strong actionability and excellent validation checkpoints, and it uses one-level-deep references effectively. Its main weakness is length and repetition, which hurts token efficiency.
Suggestions
De-duplicate rules that recur across sections: state each prohibition once in its home section and reference it from §8 completion conditions instead of restating it.
Move the long contract YAML schemas (e.g. MASTER-VISUAL-CONTRACT, DETAIL-CONTRACT) into a reference file and keep only the field list inline, reducing SKILL.md length while preserving actionability.
Consolidate the repeated image-provenance and anti-fabrication rules into a single authoritative block in §3, then point other sections to it.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body assumes Claude's intelligence (no basic-concept padding), but at ~590 lines it is noticeably long with repetition — the §8 completion conditions restate rules already detailed in earlier sections, and several prohibitions (e.g. 禁止用生成图像冒充产品截图) recur across sections, so it could be tightened. | 3 / 5 |
Actionability | Provides concrete executable artifacts — copy-paste YAML contract schemas (VISUAL-THESIS, MASTER-VISUAL-CONTRACT, decision card), specific npm packages, tool commands (visual-web-info --json, git clone --depth 1), a worked example, and five defined rework tests — but this is mixed with abstract principle statements, leaving minor gaps. | 4 / 5 |
Workflow Clarity | A clear §1–§8 sequence with explicit gated checkpoints ("不得…前", "通过后…才能"), feedback loops (failure stops implementation and preserves the raw error), and a final completion checklist — matching the anchor for clear sequence with explicit validation and error-recovery loops. | 5 / 5 |
Progressive Disclosure | References are well-signaled and one level deep ("按需读取[…](references/…)" to aesthetic-routing, tool-install, stack-routing, design-system-routing, assets-and-imagegen, decision-traces, template-pool/POOL.md, all verified real files), but the SKILL.md itself inlines substantial detail (long contract schemas, rework tests) that keeps it from being a lean overview. | 4 / 5 |
Total | 16 / 20 Passed |