Content
78%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 well-structured data-source skill: an executable quick start, concrete parameter conventions, and an excellent one-level-deep reference system where every one of ~200 interface docs exists and is organized by category. Its main weakness is token efficiency — a verbatim-duplicated overview section and a verbose description column that repeats content already present in the linked reference files — plus the absence of error-handling guidance for API limits.
Suggestions
Remove or shrink the 概述 section, which repeats the frontmatter description word-for-word and adds no new information.
Drop or drastically shorten the 描述 column in the interface table (or keep only the interface name/category); each linked reference file already contains the full description, parameters, and usage notes.
Add a brief error-handling note to 快速上手 covering token validation failure and the per-request row caps (e.g., 100/200/6000 rows) with the multi-request-by-date pattern, since the table mentions these limits but the workflow never tells the agent how to respond to them.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The quick start and parameter sections are lean with no explanation of concepts Claude already knows, but the 概述 section duplicates the frontmatter description verbatim, and the ~230-row interface table's 描述 column repeats operational notes ('数据由Tushare社区自产…可以联系管理员评估', '建议在每日19点后提取') that already live in each linked reference file. This fits anchor 3 (mostly efficient but could be tightened) — not 2 because the bulk is a navigational index rather than padded prose, not 4 because the duplicated overview and verbose table column are noticeable inefficiencies. | 3 / 5 |
Actionability | The body gives fully executable, copy-paste-ready guidance: 'pip install tushare -i https://pypi.tuna.tsinghua.edu.cn/simple', 'export TUSHARE_TOKEN=your_token', and a complete Python example with token fallback ('token = os.getenv(\'TUSHARE_TOKEN\') or ts.get_token()') plus a concrete query (pro.stock_basic with fields). Concrete parameter formats (YYYYMMDD, ts_code like 000001.SZ, pandas DataFrame return) and two runnable example scripts cover the common cases — matching anchor 5. | 5 / 5 |
Workflow Clarity | 快速上手 presents a clear ordered sequence — install, register/token, locate the interface in the docs, query with code — each with a concrete command or link. However, there are no validation/error checkpoints (e.g., verifying the token works, handling the 100/200-row-per-request caps or rate limits that the table itself mentions). This fits anchor 4 (clear sequence, minor validation gaps): above 3 because the sequence is concrete and complete, below 5 because no verification or error-recovery feedback loop is described. | 4 / 5 |
Progressive Disclosure | The SKILL.md body is a true overview (quick start, parameter formats, script links) with per-interface detail pushed exactly one level deep into references/ organized by category directories matching the 分类 column — all 196 uniquely referenced files (e.g. references/股票数据/基础数据/股票列表.md, which contains the full input/output parameter tables) were verified to exist on disk, and the example scripts exist under scripts/. This matches anchor 5 (clear overview, well-signaled one-level-deep references, easy navigation). | 5 / 5 |
Total | 17 / 20 Passed |