Lingxi Li
How I run a fleet of engineer bots that manage cloud agents, keep a feedback loop closed, and run the rest of the mini org.
What you need
- Connected apps
- Cursor cloud agents (including private workers on your own machines), Slack and a shared Notion database.
- Routines
- Every 30 minutes, each bot reviews the Notion PR database. Every morning at 5 a.m., the operations bot, Jenny, holds a 1:1 with each bot.
- Skills
- The author’s own skills, such as /lingxi-design, /react-native-best-practices, /lingxi-review and /lingxi-product.
What prompted it
Constant context-switching between cloud agents the author was managing by hand.
What Grok Bot did
- Each engineer bot owns one area of the product.
- Given a task, a bot starts a cloud agent with the author’s skills and a prompt that spells out the proof expected.
- It monitors the run, checks screenshots against the request, and pushes back on or unblocks the agent.
- Every 30 minutes it checks each PR for review findings, failing CI and merge conflicts, and sends fixes back to the agent.
- Low-risk PRs with a confident review merge automatically; the rest wait for the author.
Watch out for
- Each bot has its own memory and limited context, so the bots work best within a single domain.
- Flaky failures still reach the author when Grok Bot lacks the security permissions to fix them.
