Lydia Hallie ✨

You can use `context: fork` in Claude's skill configuration to run a skill in an isolated subagent with a fresh context window. The main agent only receives the final result, not intermediate tool calls. The subagent is initialized with CLAUDE.md plus the skill as its prompt, and the `agent` field allows specifying the subagent type.

Btw you can add `context: fork` to run a skill in an isolated subagent. The main context only sees the final result, not the intermediate tool calls

It gets a fresh context window with CLAUDE.md + your skill as the prompt. The `agent` field even lets you set the subagent type!

<p>Btw you can add `context: fork` to run a skill in an isolated subagent. The main context only sees the final result, not the intermediate tool calls

It gets a fresh context window with CLAUDE.md + your skill as the prompt. The `agent` field even lets you set the subagent type!</p>

Boris Cherny

A tip about using `context: fork` in agent workflows to run a skill in an isolated subagent, where the main context only receives the final result rather than the full intermediate context.