你是 DoneBench 的实现助手。请只读分析以下文件并给出最小实现方案，不要改文件：donebench/scripts/advanced_stats…
- Stable ID:
codex-user-prompt:3d761dcf2aedb13d
- Source kind:
codex-session-user
- Category:
research-workflow
- Timestamp:
2026-05-09T23:40:34.676Z
- Semantic hash:
3d761dcf2aedb13de2f6cabb6bd0f145b074ed899d63c63e497f220b5f6286b4
- Public handling: selected full-text prompt with secret filtering.
Prompt Text
你是 DoneBench 的实现助手。请只读分析以下文件并给出最小实现方案，不要改文件：donebench/scripts/advanced_stats.py, donebench/scripts/failure_mining.py, donebench/scripts/near_miss_breakdown.py, donebench/scripts/paper_refresh.py, donebench/scripts/experiment_pipeline.py, donebench/cli.py. 目标是生成 M6.1 诊断输出：四象限表、自我违反 taxonomy、near-miss × success 诊断表，输出到 reports/full_runs/runs/topconf_deepseek_toolplan_full/diagnostics/，并刷新 paper/tables。请返回：1) 哪个现有脚本最适合扩展；2) 需要哪些输入文件；3) 建议输出文件名；4) 最小新增代码路径。
Reuse Notes
- EXTRACTED: This is a selected Codex prompt or automation prompt from the local Codex corpus.
- INFERRED: Future agents can reuse its structure, constraints, and acceptance criteria when creating similar Codex workflows.