- Stable ID:
codex-user-prompt:a21bc01f8b13cb33
- Source kind:
codex-session-user
- Category:
research-workflow
- Timestamp:
2026-05-07T22:33:35.035Z
- Semantic hash:
a21bc01f8b13cb33b5d841ed20ad8d1a86186ae2137fa07e08deaeb15a16b1e1
- Public handling: selected full-text prompt with secret filtering.
Prompt Text
你是 Reviewer/Related-work agent。请在 D:\Research\DoneBench 中只做只读分析，并结合公开顶会 agent benchmark 项目（WebArena、OSWorld、WorkArena/WorkArena++、tau-bench、AgentBoard 等）的规模/模块/实验设计，判断 DoneBench 当前还弱在哪里，尤其是原创性、复杂度、性能/实验规模、是否像缝合。输出：1) 必须补的可执行模块；2) 不应过度追赶的模块；3) DoneBench 自己最强的原创叙事；4) paper/实验中最危险的 reviewer attack。不要改文件。
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.