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 a strong operational document: exact commands, exit codes, a QR-login flow with explicit stop-and-wait gates and recovery loops, wujie shadow-DOM gotchas, and concrete pitfalls. Its main weaknesses are heavy verbatim repetition of the same session/cookie constraints across four sections and a complete failure to reference the bundled implementation script (scripts/fetch_engagement.py), leaving the CLI-to-code mapping implicit.
Suggestions
Reference the implementation explicitly in the CLI section, e.g. 「实现：scripts/fetch_engagement.py（python3 scripts/fetch_engagement.py <subcommand> …）」, so the commands map to the actual bundle file
Consolidate the repeated session constraints (不导出 cookie/UA/token, fail-first 队列, --session wechat-channel --persistent) into the single Session 共享约束 section and reference it from elsewhere instead of restating each 3-4 times
Specify what 「归一化匹配」 means for matching row.title against the 作品管理页 descriptions (e.g. whitespace/case/emoji normalization) so step 5 of the fetch flow is reproducible
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational with no explanations of concepts Claude already knows, but core constraints are restated 3-4 times: 「不导出 cookie/UA/token」 appears in the Session 共享约束, 抓取流程 notes 2 and 3, and 约束 sections; 「fail-first 队列…不自动 close」 appears in four places (lines 26, 101, 161, 185-188); 「--session wechat-channel --persistent」 is repeated at lines 36, 101, 119, and 200. This matches anchor 3 ("mostly efficient but could be tightened"): not 4 because the duplication is substantial rather than minor; not 2 because there is no padding or tutorial-style explanation. | 3 / 5 |
Actionability | Concrete, executable guidance throughout: exact CLI commands with arguments (「wx-channel-engagement fetch --row-id <id>」), an exit-code table (0/1/2/3), exact URLs, a copy-paste eval snippet (「document.querySelector('wujie-app').shadowRoot.innerText」), the innerText structure template, and a full output JSON example. Anchor 4 ("mostly executable guidance... with minor gaps"): not 5 because the body never states how the `wx-channel-engagement` commands map to the bundled `scripts/fetch_engagement.py`, and 「按完整视频描述归一化匹配」 does not define the normalization; not 3 because the commands are real and executable, not pseudocode. | 4 / 5 |
Workflow Clarity | The fetch flow has explicit validation branches (「redirect URL 跳 login/扫码页 → exit 2」) and a guaranteed cleanup step (「finally: close session」), and the login flow has a hard user-confirmation gate (「Stop and wait：等用户回复…再往下走，不盲轮询」) plus a full error-recovery loop (「exit 2 = 未就位 → 提示用户点确认后重跑；已确认仍未就位 → 重跑 login 生成新二维码」). The 后台 DOM 改版 pitfall supplies a probe → inspect dumps → adjust-parser feedback loop, and the 报错约束 mandates verbatim stderr + exit code reporting. This matches anchor 5 ("clear sequence with explicit validation steps; feedback loops for error recovery"); not 4 because checkpoints and recovery branches are explicit at every fragile step rather than mostly present. | 5 / 5 |
Progressive Disclosure | The bundle contains one implementation file (scripts/fetch_engagement.py) yet the body never references, links, or names it — the CLI section presents `wx-channel-engagement` commands with no pointer to where the code lives, so an agent cannot discover the script from the document. In-body sectioning itself is good (numbered sections, pitfall entries, exit-code table), which fits anchor 3 ("some structure but could be better organized" with references not clearly signaled — here the bundle reference is entirely absent): not 4 because a provided bundle file is completely unsignaled; not 2 because the body is well-sectioned rather than a minimal monolith. | 3 / 5 |
Total | 15 / 20 Passed |