让 Claude 接管应用日常维护:388 个 PR 的实测
作者近几周做了一个实验:让 Claude 在独立的 Slack 频道里接管 iOS、Android、Desktop、web、CLI 与 Agent SDK 的日常维护。Claude Tag 每天执行崩溃 fuzzer、重复抽象统一、死代码清理、抽象纪律检查等 routine,先自动开 PR,再经 Claude Code Review 和人工复核合并。几周内在各个仓库开出 388 个 PR,其中 180 个被合并。Claude 大多一次通过,出错时作者会让 Claude 调校自身 routines,往往几天内见效。适合正在探索 agent 接管机械性维护工作的工程团队参考。
A weird experiment I've been trying the last few weeks is having Claude take over day-to-day maintenance of our apps. Seeing early signs of life that this might be possible.
过去几周,我一直在做一个有点奇怪的实验:让 Claude 接管我们应用的日常维护。目前已经看到一些早期迹象,表明这件事或许真的可行。
The setup is straightforward: we have a Slack channel called proj-claude-maintains-apps. In it, Claude Tag runs a bunch of daily routines across iOS, Android, Desktop, web, CLI, and Agent SDK:
- Crash fuzzer: open the app in a simulator and tap around to find ways to crash it, then root cause and fix the crashes
- Dup unifier: scans the codebase for similar-yet-slightly-divergent abstractions, and puts up PRs to unify them
- Dead-code remover: removes statically unreachable code, and adds logging to suspected dead code to check if it's really dead and if so, remove it the next day
- Abstraction police: fixes leaky abstractions
- a bunch more..
配置很简单:我们建了一个叫 proj-claude-maintains-apps 的 Slack 频道。Claude Tag 会在里面跑一批日常任务,覆盖 iOS、Android、Desktop、web、CLI 和 Agent SDK:
- 崩溃模糊测试(Crash fuzzer):在模拟器里打开应用,随机点按以寻找崩溃路径,然后做根因分析并修复。
- 重复代码统一(Dup unifier):扫描代码库中相似但略有分歧的抽象,提交 PR 将它们统一。
- 死代码清理(Dead-code remover):移除静态不可达代码;对疑似死代码加上日志,确认真的没人用后,第二天再删除。
- 抽象警察(Abstraction police):修复泄漏的抽象。
- 还有不少其他任务……
Results have been surprisingly positive. Over the last few weeks, these routines have opened 388 PRs across our repos, 180 of which we merged after Claude Code Review + human review. We're now thinking about how to streamline this to make merging these kinds of mechanical changes easier.
结果出人意料地好。过去几周,这些例行任务在我们的仓库里累计开了 388 个 PR,其中 180 个在 Claude Code Review 和人工 review 后被合并。我们现在正在思考如何进一步优化流程,让这类机械性变更更容易合并。
Claude generally gets these PRs right on the first shot, and if it doesn't, we ask Claude to tune its routines so it's better the next day. Sometimes it takes a few days of tuning.
Claude 通常第一轮就能把 PR 做对;如果没做对,我们会让它调整例行任务,第二天做得更好。有时候需要调上好几天。
To try a similar workflow, ask Claude Code or Tag, or create some routines directly at https://claude.ai/code/routines. A few of the actual prompts I used below.
Has anyone experimented with similar workflows?
想试试类似的工作流,可以让 Claude Code 或 Tag 来跑例行任务,也可以直接到 https://claude.ai/code/routines 创建一些例行任务。下面是我实际用过的一些提示词。
有人做过类似的实验吗?



