DoorDash 用多智能体 LLM 自动清理 6 万个功能开关
原标题：DoorDash Uses Multi Agent LLMs to Clean up 60,000 Feature Flags
AI 摘要
DoorDash 构建了一套多智能体 LLM 系统，用于自动清理代码库中过期的功能开关（feature flag）。该系统结合实时实验数据、工程师审批、隔离的 Git worktree 和自动验证，在 50 个过期开关的评估中为 45 个生成了可用的 pull request，平均每次清理耗时 13.8 分钟、成本 4.79 美元，而人工清理估计需要 1 到 2 小时。系统使用 Google 的 Agent Development Kit，由 Claude Sonnet 编排、Claude Opus 执行清理，并通过 MCP 查询实验平台元数据。该工作已被 ICSME 2026 工业赛道接收。
正文节选
DoorDash has built a multi-agent LLM system to automate stale feature flag cleanup across its codebase, combining live experimentation data, engineer approval, isolated Git worktrees, and automated validation. In an evaluation of 50 stale flags, the system produced usable pull requests for 45, averaging 13.8 minutes and $4.79 per cleanup, compared with DoorDash’s estimate of one to two hours for manual cleanup. DoorDash’s experimentation platform manages more than 60,000 feature flags across rou