Content
92%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.
A tightly written, highly actionable skill body: exact commands, output conventions, an explicit data-quality feedback loop, and clean one-level-deep reference structure. The only trims available are a few justificatory sentences in the philosophy section and Phase 4 caveats.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes Claude's competence — no basic-concept explanations, platform detail deferred to references — but the 哲学 rationale sentences (仅凭题材名无法判断具体情绪体验) and the long Phase 4 caveat paragraph contain minor over-explanation that could be trimmed, so it sits just below level 5. | 4 / 5 |
Actionability | Fully executable guidance: exact aggregation commands with flags (--sparse 10 --scale short, --sample … --n 5), concrete output-directory conventions (扫榜/{YYYYMMDD}/ with -2 suffix), verbatim user-facing question text, an emotion mapping table, and a complete copy-paste report template. | 5 / 5 |
Workflow Clarity | Five clearly sequenced phases with an explicit validation feedback loop — check the aggregate's 「采集情况」, re-scrape via the platform reference when problems are flagged, and disclose exclusions in the report when recovery fails — plus honesty rules like omitting the 变化 section for single-period samples. | 5 / 5 |
Progressive Disclosure | The body is a well-signaled overview: platform-specific detail is split into one-level-deep references loaded selectively (只读所选平台那份), a 参考资料 table gives per-file load timing (何时加载), and all referenced files (platform-dianzhong.md, platform-heiyan.md, manual-rank-input.md, real-market-data.md, topic-match.md) exist with no nested references. | 5 / 5 |
Total | 19 / 20 Passed |