Content
92%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 skill body: concrete search/output specs, a mandatory numeric self-consistency check with a worked example, and a validation-gated workflow. The only trimmable material is the list of well-known theory frameworks and a stray reference to a user memory.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is largely lean — search templates, a worked CAGR check, and a checklist with no padding — but the 8-entry theory list (e.g., "迈克尔·波特《竞争战略》— 五力模型、价值链") explains frameworks Claude already knows, and the "见用户 [记忆: web 搜索验证财务事实]" line is leaky/unnecessary. These are minor trimmable spots, matching the anchor-4 'efficient; minor instances of over-explanation'. | 4 / 5 |
Actionability | Fully executable guidance throughout: a literal 8-query search template code block, exact output path "markdown/{行业名}-行业深度分析.md" (with directory-creation rule), a length spec of 6000-12000 字, the concrete formula "首值 × (1 + r)^n = 末值" with a worked numeric example (970 × 1.38^5 = 5238 亿 vs 730), a 6-question entry-strategy checklist, and a pre-writing self-check list. | 5 / 5 |
Workflow Clarity | A clearly numbered 5-step workflow with explicit validation and feedback loops: Step 2 mandates live search verification, the data-consistency check is labeled "硬性约束, 落地前必做" with "发现矛盾立即重查原始来源" (verify → fix → retry), and Step 5 is gated by a 10-item pre-output checklist. | 5 / 5 |
Progressive Disclosure | SKILL.md stays an overview and pushes detail to four one-level-deep, real bundle files (framework.md, output-template.md, visualization-guide.md, data-sources.md — all present in references/), each linked at its point of use and summarized again in a "资源文件" section; nothing that belongs in a reference is inlined. | 5 / 5 |
Total | 19 / 20 Passed |