Content
65%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 body is a high-value, actionable regulatory reference with concrete numbers and a useful compliance checklist, but it is monolithic — all content is inlined with no reference files — and lacks explicit validation checkpoints in its analysis workflow.
Suggestions
Move per-market detail (港股/美股/加密 sections) into separate reference files (e.g. references/hk-us-crypto.md) and keep SKILL.md as an overview with clearly signaled one-level-deep links, improving progressive disclosure.
Add explicit validation checkpoints to the analysis framework (e.g. 'Verify each market's T+N and 涨跌停 assumptions against the current-date rule set before running the backtest') to raise workflow clarity.
Tighten repeated '回测影响' preamble lines and convert illustrative pseudocode blocks into concrete runnable snippets or clearly marked checklists.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense reference content where most lines carry specific rule data Claude would not reliably know (limit percentages, T+N, fee rates, PDT thresholds, tax rates); assumes competence with only minor repeated '回测影响' framing that could be trimmed. | 4 / 5 |
Actionability | Concrete numeric tables, fee breakdowns, time windows, PDT/熔断 thresholds, a checkbox compliance checklist, and a full example output report give mostly executable guidance; a few code blocks are illustrative pseudocode rather than runnable code. | 4 / 5 |
Workflow Clarity | An 分析框架 with a compliance checklist and a labeled output template provides a rough sequence, but validation checkpoints and error-recovery feedback loops are absent or only implicit. | 3 / 5 |
Progressive Disclosure | No bundle/reference files exist and all ~300 lines are inlined in SKILL.md; header structure is decent, but market-specific detail that could live in separate reference files is monolithic rather than split with one-level-deep navigation. | 3 / 5 |
Total | 14 / 20 Passed |