In my original project, I now have 300+ FD files each with a problem statement, solutions considered, and what was implemented. An emergent property of this system is that agents frequently rediscover past FDs on their own with /fd-explore /fd-deep or when they launch Explore agents with plan mode. The added context of what was considered prior helps the agents plan better, and also remind me of relevant work I may have forgotten about (with all the context switching between tabs and the increased speed of work I am less capable of remembering what I have done).
System.out.println("ForkJoinPool : 线程阻塞 → tryCompensate → 唤醒/创建线程维持并行度");，这一点在新收录的资料中也有详细论述
，推荐阅读新收录的资料获取更多信息
编者按：本文是少数派 2025 年度征文活动#TeamCarbon25标签下的入围文章。本文仅代表作者本人观点，少数派只略微调整排版。。新收录的资料对此有专业解读
Agent 指挥 Agent —— 专为自动化编排设计，CLI 完全自描述，任何具备 shell 执行能力的 Agent 都能自主驱动