Content
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is highly actionable with concrete interfaces, formulas, a real bundled script, and a well-sequenced validated workflow. Its weaknesses are verbosity (redundant prohibition/interface sections) and a monolithic structure that underuses progressive disclosure.
Suggestions
Remove redundancy: drop the standalone '禁止事项' list (fold unique items into their relevant steps) and merge the '依赖的 tushare 接口清单' into the data-need mapping table to tighten conciseness.
Extract the large threshold tables and data-need mapping into a separate REFERENCE.md (one level deep, clearly linked) to improve progressive_disclosure.
Trim the '系数说明'/'为什么要这条规则' prose where the rule itself is self-explanatory to reduce token load.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient domain-specific guidance, but contains redundancy — the '禁止事项' list reiterates earlier rules and the '依赖的 tushare 接口清单' duplicates the data-need mapping table — that could be tightened. | 2 / 3 |
Actionability | Provides concrete tushare/akshare interface names with parameters, an explicit dividend formula ('cash_div_tax / 10 × total_share'), executable bash commands for the sina script, and a full end-to-end pseudocode skeleton that is copy-paste ready. | 3 / 3 |
Workflow Clarity | Sequences a clear M0 → Step1–6 flow with explicit validation checkpoints (akshare fallback labeling, dividend dedup, '禁止' rules) and a mandatory dual-axis output template that acts as a completion checklist. | 3 / 3 |
Progressive Disclosure | The one bundled script is correctly referenced (scripts/fetch_sina_penalties.py exists and is linked), but the body is a monolithic ~500-line wall with threshold tables, formulas, and interface lists kept inline rather than split into one-level-deep reference files. | 2 / 3 |
Total | 10 / 12 Passed |