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.
The body is a well-structured, genuinely actionable index: an executable quick start, a categorized table routing to 229 real one-level-deep reference documents, and separated example scripts. The main weaknesses are redundancy — the overview duplicates the description verbatim and many table descriptions restate their titles — and a broken import in the example script that keeps it from being copy-paste runnable.
Suggestions
Delete the 概述 section (it duplicates the frontmatter description word-for-word) and trim table descriptions that merely restate the title, keeping only rows where the description adds information such as update cadence, record limits, or history coverage (conciseness).
Fix `scripts/stock_data_example.py` and `scripts/fund_data_example.py` to read the token via `os.getenv('TUSHARE_TOKEN') or ts.get_token()` like the inline example, instead of the missing `src.config.accessor` import, so the linked examples are executable (actionability).
Add a short '常见错误' (common errors) note covering the main failure modes — token missing/invalid, 单次最大行数 limits requiring date-windowed retries, and permission errors — to give the workflow an explicit feedback loop (workflow_clarity).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 概述 section repeats the frontmatter description verbatim, and many interface-table descriptions merely restate the row title ('可转债基础信息' → '获取可转债基本信息'), adding hundreds of redundant tokens. The quick-start and parameter-format sections are tight, so it is 'mostly efficient but could be tightened' (anchor 3) rather than noticeably padded (2) or clean (4). | 3 / 5 |
Actionability | The quick start gives an executable pip install, a token environment-variable setup, and a complete runnable `pro.stock_basic` snippet, and every interface row links to a reference file with full input/output parameter tables. Not level 5 because `scripts/stock_data_example.py` imports `from src.config.accessor import get_env_config`, a module not present in the bundle, so the linked examples are not copy-paste runnable. | 4 / 5 |
Workflow Clarity | A clear, ordered sequence (install → register and set TUSHARE_TOKEN → find the interface in the docs → query with Python) with concrete commands at each step, and the example scripts include try/except error reporting. Not level 5 because the body offers no explicit feedback-loop guidance (e.g. what to do on rate-limit or permission errors), though for read-only queries the missing checkpoints are minor. | 4 / 5 |
Progressive Disclosure | Textbook hub-and-spoke structure: SKILL.md is an overview plus a categorized index, all per-interface detail is split into one-level-deep `references/` files (verified present on disk), and runnable-length examples live in `scripts/`. Each row links directly to its detail file, so navigation is unambiguous and nothing is nested more than one level deep. | 5 / 5 |
Total | 16 / 20 Passed |