Content
68%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 dense, information-rich reference skill: nearly every section delivers non-obvious domain specifics (quantified thresholds, weighted scorecards, an output template) with little padding. Its weaknesses are structural — the analysis workflow is never sequenced explicitly with validation checkpoints, the verification code is illustrative rather than executable, and the single ~260-line file inlines reference tables that belong in bundle files.
Suggestions
Turn the implicit process into an explicit numbered workflow (e.g. 1. 拉取数据 via tushare → 2. 勾稽验证 → 3. 盈利质量评分 → 4. 红旗检查 → 5. 杜邦分解 → 6. 按模板输出), embedding the 勾稽验证公式 as validation checkpoints at each step.
Make the verification snippets executable: show the actual data fetch (tushare pro API calls per the 注意事项 #6) and a runnable function for accrual_ratio / receivable-vs-revenue growth, instead of fragments referencing undefined variables.
Move the 12-red-flag table, the industry ROE comparison, and the full output template into a references/ file (e.g. references/red-flags.md), keeping SKILL.md as a tighter overview with clearly signaled links.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and table-driven: nearly every line carries non-obvious domain data (thresholds like '存贷双高 >营收30%', the CFO/CFI/CFF quality matrix, weighted scoring cards, industry ROE profiles) rather than explaining what Claude already knows. Minor over-explanation exists — '资产 = 负债 + 所有者权益' and the textbook ROE three-factor identity are standard knowledge — placing it at 'efficient; minor instances that could be trimmed' rather than the fully lean 5. | 4 / 5 |
Actionability | Most guidance is concretely executable: quantified red-flag detection rules (e.g. '应收增速 > 营收增速 × 1.5，持续2季+'), a weighted earnings-quality scorecard with cutoffs (≥2.5 / 1.5-2.5 / <1.5), and a full markdown output template. It stays below 5 because the Python verification snippets are illustrative fragments (undefined variables like net_income, cfo; no data-fetching shown despite tushare being mentioned), and below the anchor's 'copy-paste ready' bar. | 4 / 5 |
Workflow Clarity | The output-format template implies a sequence (三表概要 → 盈利质量评分 → 杜邦分解 → 红旗检查 → 结论), and the 勾稽验证公式 act as validation checks, but the body is organized as parallel knowledge sections rather than an ordered analysis workflow — the checks are never wired into the steps as explicit checkpoints. This matches 'sequence present but checkpoints missing or implicit', and it is not a destructive/batch operation so the hard cap of 3 also fits naturally. | 3 / 5 |
Progressive Disclosure | No bundle files exist, so everything is appropriately self-contained in one file with clean, well-signaled section headers (三表核心框架 → 勾稽关系 → 盈利质量 → 红旗 → 杜邦 → 输出格式 → 注意事项) that are easy to navigate. Not 5 because at ~260 lines the 12-red-flag table, industry ROE comparison, and output template are reference material that would serve better in a split-out references/ file, keeping the main file as an overview. | 4 / 5 |
Total | 15 / 20 Passed |