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 dense, well-structured reference with genuinely specific A股 domain data, quantified thresholds, and clear multi-step frameworks. Its weaknesses are the absence of any progressive-disclosure file split (~270 lines all in SKILL.md) and the lack of executable code for the core computations.
Suggestions
Split heavy reference material into one-level-deep bundle files (e.g. references/etf-comparison.md, references/style-drift.md, references/manager-evaluation.md) and keep SKILL.md as a concise overview with clearly signaled links — this is the lowest-scoring dimension.
Add a short runnable snippet for the core computations (rolling-window style regression and T-M model) instead of only math formulas, since the skill already declares pandas/numpy/scipy dependencies.
Add explicit validation/feedback checkpoints to the workflows, e.g. what to re-check when R² < 0.70 or when tracking error exceeds the 2%/4% thresholds, to move workflow clarity from a static checklist to a validate-and-adjust loop.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and token-efficient — threshold tables, formulas, and A股-specific index codes (e.g. "沪深300价值 (399346)") with almost no prose padding or explanation of concepts Claude already knows — but the sample ETF comparison table and illustrative report could be trimmed or split out, keeping it below the lean anchor of 5. | 4 / 5 |
Actionability | Concrete, quantified guidance throughout ("|Δβ| > 0.2 → 显著漂移", "R² > 0.85 → 风格明确", "日均成交额 > 1亿", per-step checklists, and an output report template), but executable guidance is limited: the only runnable code is "pip install pandas numpy scipy", and the core regression/T-M model computations are given as formulas rather than runnable snippets — minor gaps consistent with score 4. | 4 / 5 |
Workflow Clarity | The five-step screening framework (硬指标过滤 → 绩效排序 → 风格验证 → 基金经理评价 → 费用检查) and four-step FOF construction are clearly sequenced with quantified checkpoints ("任一资产偏离目标权重 > 5% → 再平衡") and monitoring rules. This is an analytical skill with no destructive or batch operations, so the 3-cap does not apply; a 5 would require explicit validate→fix→retry feedback loops, which are absent. | 4 / 5 |
Progressive Disclosure | The ~270-line body is well-sectioned with clear headers, but it is a single monolithic file with no bundle files at all — content that would fit separate references (the ETF comparison table, style-drift detection details, the full manager-evaluation framework, the sample report) is fully inlined. That sits above anchor 2 (which lacks organization) and below anchor 4 (which splits content appropriately across files). | 3 / 5 |
Total | 15 / 20 Passed |