Content
70%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 operationally excellent: complete CLI syntax, option tables, output schemas, and batch workflows with genuine validation checkpoints and failure-handling loops. Its weaknesses are redundancy (usage examples and follow-constraint restated multiple times) and a missing bridge to the bundled script that implements everything, which also leaves the invocation path implicit.
Suggestions
Add a line connecting the CLI to the bundle, e.g. 'run via python3 scripts/wx_mp_hunter.py <command>' (or document the install/PATH setup), so the copy-paste commands are actually executable from the skill bundle.
Remove the '典型用法示例' section or reduce it to a single compact table — it duplicates the workflow diagram and per-command sections already shown above.
State the posts-list follow-account constraint once and reference it from the JSON examples' hint field instead of repeating the full explanation in three places; consider moving detailed environment-detection paths into a short reference file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The '典型用法示例' section (scenarios A/B/B2/C) largely restates the '两条独立工作流' diagram and the per-command sections, and the posts-list follow-account constraint is repeated three times (workflow note, 已关注账号约束 section, and both JSON hint examples). This is noticeable tightening opportunity rather than minor trims, so anchor 3 fits better than 4, though it never explains concepts Claude already knows. | 3 / 5 |
Actionability | Commands are concrete and copy-paste ready with full flag syntax ('wx-mp-hunter fetch <url> --html --download-images --output-dir ./article-out'), option tables, and annotated output JSON with field-by-field descriptions. It falls short of anchor 5 because the body never references the bundled scripts/wx_mp_hunter.py, leaving unstated how the 'wx-mp-hunter' command maps to or invokes the script, and camoufox-cli steps are described procedurally rather than with executable commands. | 4 / 5 |
Workflow Clarity | The homepage and batch-fetch workflows have explicit validation and feedback loops: small-sample verification before batch ('先选 1 篇文章 fetch 验证链路成功；成功后才能批量'), scroll-until-height-stable checks, URL dedup with count reporting, 1-2s pacing, stop after 3 consecutive failures, captcha escalation to the user, and an error-cause-action table. This matches the anchor 5 pattern of clear sequence with explicit validation and error recovery. | 5 / 5 |
Progressive Disclosure | The body has clear per-command sections, but the bundle's only implementation file (scripts/wx_mp_hunter.py) is never referenced or linked from the body, and ~300 lines of inline CLI reference (options tables, two full posts-list output examples, detection path details) sit in SKILL.md with no one-level-deep split. Structure exists but organization and signaling gaps keep it at anchor 3 rather than 4. | 3 / 5 |
Total | 15 / 20 Passed |