Content
50%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A dense, domain-specific analytical skill with concrete formulas and a structured report template, but it is a monolithic single file with illustrative rather than executable code and no explicit validation checkpoints. Splitting event types into references and adding runnable examples would lift it.
Suggestions
Add at least one fully executable example (e.g., a tushare snippet fetching 增减持/定增 data) since tushare is already cited as the data source, replacing the purely illustrative calculation blocks.
Split each event type (并购套利, 增减持, 股权激励, 定增/可交债, ST/退市) into one-level-deep reference files (e.g., references/merger-arb.md) linked from a concise overview, instead of keeping everything inline.
Add explicit validation/checkpoint steps to the workflow (e.g., confirm the announcement source + apply the 过滤规则 before sizing a position), and move dated/time-sensitive figures into a labeled section noting their sourcing.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient and adds A-share-specific knowledge Claude lacks (regulatory thresholds, timelines, signal tables), but restated prose in 概述/注意事项 and time-sensitive dated items ('2024新规', '2023年某科技公司') outside a deprecated/old-patterns section keep it from the fully lean score-3 anchor. | 2 / 3 |
Actionability | It gives concrete formulas, thresholds, and a copy-paste report template, but the calculation blocks are illustrative rather than executable and tushare is named as a data source without any runnable fetching code, fitting the score-2 'incomplete / pseudocode' anchor rather than 'fully executable'. | 2 / 3 |
Workflow Clarity | Event timelines (预案→股东大会→证监会→发行→解禁) and T-30/T/T+1 windows are clearly sequenced with gating filter rules, but there is no explicit validate→fix→retry checkpoint, so it sits at the score-2 'sequence present but checkpoints implicit' anchor. | 2 / 3 |
Progressive Disclosure | The ~260-line document is well-organized into sections but is monolithic with no bundle/reference files, and content that could be split per event type sits inline; this matches score-2 'some structure but content that should be separate is inline' rather than score-3 split-with-references. | 2 / 3 |
Total | 8 / 12 Passed |