Content
63%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 well-structured, actionable multi-step research workflow with explicit checkpoints and a quality checklist. Its main weaknesses are padding in the writing-style section and an orphaned reference file (references/schema.json) that is never linked from the body.
Suggestions
Condense the '写作风格' section into a short checklist of do/don't elements rather than re-explaining each rhetorical device at length; rely on Claude's competence to reduce token load (improves conciseness).
Reference the bundled output schema from the body — e.g. '报告字段结构见 references/schema.json' — so the report-output schema is discovered via progressive disclosure instead of remaining an orphaned file (improves progressive_disclosure).
Add an explicit validation checkpoint before PDF generation — e.g. 'Verify the first-line H1 and the 研究时间/所属领域/研究对象类型 meta line are present, and re-run md_to_pdf.py if WeasyPrint reports errors' — so the PDF step has a clear feedback loop (improves workflow_clarity).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient task guidance, but the '写作风格' section re-explains style elements ('节奏感','层层剥开的修辞','文化升维','回环呼应') at length and the prose is discursive ('宁可写长写细，也不要蜻蜓点水'), adding padding beyond what Claude needs. Not 4 because the over-explanation is more than minor; not 2 because it does not explain basics Claude already knows and is not severely verbose. | 3 / 5 |
Actionability | Provides concrete, copy-paste-ready commands ('pip install weasyprint markdown --break-system-packages','python [skill目录]/scripts/md_to_pdf.py input.md output.pdf --title … --author …', arxiv 'curl -s "https://export.arxiv.org/api/query?search_query=…"') covering common cases. Not 5 because sub-agent search prompts are described in prose rather than given as ready-to-paste templates; not 3 because the guidance is executable, not pseudocode. | 4 / 5 |
Workflow Clarity | Clear 5-step sequence with explicit checkpoints ('信息充分性自检…信息不够就再补搜', a '质检清单' with checkboxes) and a re-search feedback loop; no destructive/batch cap applies since output is a generated report. Not 5 because there is no explicit validate-the-PDF-output step or recovery path if WeasyPrint fails; not 3 because checkpoints are explicit rather than implicit. | 4 / 5 |
Progressive Disclosure | Good section structure and a clearly signaled script reference ('scripts/md_to_pdf.py'), but the bundled 'references/schema.json' (~7.8KB report-output schema) is never referenced from the body, and report-structure content that could point to it is inlined. Not 4 because the unreferenced bundle file is a real organization gap; not 2 because structure is reasonable, not minimal or buried. | 3 / 5 |
Total | 14 / 20 Passed |