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 an efficient, well-structured overview with concrete commands, formats, and a sound escalation workflow. Its main defects are the four dangling templates/*.py references that undermine both navigation and the script-generation step, and the absence of an explicit result-validation checkpoint.
Suggestions
Add the four missing template files (templates/basic_fetch.py, templates/stealth_cloudflare.py, templates/session_login.py, templates/parse_only.py) referenced by the decision tree and template index, or replace the template mechanism with inline copy-paste-ready code examples in the body or a real reference file.
Insert an explicit validation checkpoint between "执行脚本" and "返回结果" in the workflow — e.g., verify status code, non-empty content, and expected selectors before reporting success, especially for batch (批量采集) runs.
Consider narrowing overly generic trigger terms in the description ("fetch page", "extract data") to reduce overlap with general web-fetch skills.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence: a version-check command, a compact decision tree, and dense quick-reference tables for cookie formats and timeout units, with zero padding or explanation of concepts Claude already knows. Every section earns its tokens, matching the score-5 anchor. | 5 / 5 |
Actionability | Concrete executable guidance throughout — exact commands ("pip install \"scrapling[fetchers]\"", "scrapling install"), exact cookie dict/list[dict] formats with examples, and per-fetcher timeout units. It falls short of score 5 because the core step "读取对应模板 → 替换参数 → 生成完整脚本" depends on templates/*.py files that do not exist in the bundle, and no inline code examples backstop them. | 4 / 5 |
Workflow Clarity | A clear sequence (version check → consult site-patterns → decision tree → generate script → execute → mandatory knowledge sedimentation) with an escalation feedback loop ("先用 Fetcher 试，403/空内容 → 升级到 StealthyFetcher") and a troubleshooting reference on error. Score 5 is not reached because there is no explicit validation step verifying scrape results (e.g., non-empty content) between execution and returning results, which matters for the 批量采集/batch use case. | 4 / 5 |
Progressive Disclosure | The references index is well designed — five real references/*.md files, one level deep, each with an explicit "何时读取" trigger — but four of the nine referenced paths (templates/basic_fetch.py, stealth_cloudflare.py, session_login.py, parse_only.py) do not exist in the bundle. Since the primary workflow (generate script from template) routes through these missing files, navigation is broken for the core path, which is more than the 'minor organization gaps' allowed at score 4. | 3 / 5 |
Total | 16 / 20 Passed |