Run Copilot CLI in loops and split multi-part work

Daniel Ward used GitHub Copilot’s autopilot and /fleet features to finish larger jobs and divide separate chunks across parallel agents.

Done with GitHub Copilot· Intermediate

What they did
Ward turned on /experimental in GitHub Copilot CLI and then let autopilot handle a task with repeated passes instead of fresh prompts each time. He paired it with /plan to set direction, and used /fleet when the work broke into distinct pieces that could be handed to several subagents at once, including custom ones.
How it went
That setup let Copilot carry a substantial task through to the end, which worked well for quick proof-of-concept work. Ward found that /fleet adds coordination cost and can slow down smaller jobs, and he sets --max-autopilot-continues when he can’t monitor the session.
Tips
Use autopilot for work you can review later, not for production changes where closer oversight matters. If a run might go unattended, cap the number of automatic follow-up loops.

Try it yourself with GitHub Copilot

In [repo path], use GitHub Copilot CLI to tackle [task]. Begin with a short plan, then hand execution to autopilot and let it continue until [completion condition]. If the work naturally separates into independent chunks such as [chunk 1] and [chunk 2], route them through /fleet and spread them across parallel subagents; otherwise keep the job with one agent. Before any irreversible step like sending, deleting, paying, or submitting, pause and ask for my approval. When you’re finished, give a concise recap and note anything uncertain.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on daninacan.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.