Content
42%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 well-structured with concrete data-source paths and a clear output format, but it is padded with non-executable pseudocode stubs and explanations of standard statistics, and it references detail files that do not exist in the bundle. Tightening the stubs and providing the referenced files would substantially improve it.
Suggestions
Replace the pseudocode stubs (pearsonCorrelation, detectChangePoints, assessRisks, etc.) with executable code, or remove them and keep only the concrete data-source reading and output-format guidance.
Either create the referenced files (data-sources.md, algorithms.md, examples.md) in a references/ directory or remove the dangling '请参考' pointers and inline the essential content directly.
Add explicit validation checkpoints inside the 8-step workflow (e.g. verify data was loaded and spans enough days before running trend analysis) rather than relying only on the separate error-handling section.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Noticeably verbose: pseudocode stubs whose bodies are only comments ('pearsonCorrelation', 'detectChangePoints', 'assessRisks'), a '分析算法' section explaining standard statistics Claude already knows (CUSUM, Pearson/Spearman), and a ~70-line text-report mockup all pad the document; padding spans multiple sections rather than isolated over-explanation, keeping it at 2 rather than 3. | 2 / 5 |
Actionability | Data-source paths and the output format are concrete, but the core analysis code is non-executable pseudocode/stubs using fake helpers ('readFile', 'readAllJson') that are not among the declared allowed tools (Read, Grep, Glob, Write); the stubs keep it below 4, while the real data-source and output guidance keeps it above 2. | 3 / 5 |
Workflow Clarity | A clear 8-step sequence is present, but the batch read of many data files has no inline validation checkpoints; the separate error-handling section provides fallbacks but no explicit verify-then-proceed gates, so per the batch-operation cap it stays at 3. | 3 / 5 |
Progressive Disclosure | Section structure is good and references are signaled ('详细算法说明请参考：algorithms.md'), but no bundle files exist — the referenced data-sources.md, algorithms.md, and examples.md are dangling — and large content that belongs in separate files (algorithms, full output example) is inlined; structure keeps it above 2 but the missing files keep it below 4. | 3 / 5 |
Total | 11 / 20 Passed |