Content
80%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a dense, actionable framework with executable code and concrete scoring rules, but it presents an implicit rather than explicitly-sequenced workflow and keeps all material inline instead of progressively disclosing detail to reference files.
Suggestions
Add an explicit ordered workflow with checkpoints — e.g. 1) compute景气度 score, 2) rank综合动量, 3) cross-check产业链传导 direction, 4) compare估值/盈利/资金流, 5) emit配置建议 — plus a validation step confirming景气度 and momentum rankings agree before marking a sector超配.
Split the large景气度 scoring rubric and产业链传导 example tables into a references/ file (e.g. references/SCORING.md) and link from a concise overview so deeper material is one level deep rather than inline.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a lean reference framework of tables, code, and rules that assumes domain competence — it never over-explains PE/PB/ROE or the申万 classification — so every section earns its tokens; not the level-2 case that pads with unnecessary explanation. | 3 / 3 |
Actionability | It provides an executable sector_momentum function, explicit景气度 scoring thresholds (">30%=30分, 15-30%=22分"), fixed weight formulas, and a copy-ready output template, meeting the fully-executable anchor rather than the pseudocode level-2 case. | 3 / 3 |
Workflow Clarity | The four-dimension analysis sequence is present but only implicit (no numbered "1. score景气度 → 2. rank momentum → ..." ordering) and there are no validation/verification checkpoints, matching the level-2 anchor of steps present but checkpoints missing; not 1 because the framework structure is clear. | 2 / 3 |
Progressive Disclosure | The content is well-sectioned but is a ~190-line monolith with all detail inline and no reference files; the under-50-line exception does not apply, so it sits at the level-2 anchor of "content that should be separate is inline" rather than a clear overview with one-level-deep references. | 2 / 3 |
Total | 10 / 12 Passed |