forked-skill

You can spawn a forked sub-agent to run testing workflows in an isolated context, ensuring safe experimentation without affecting the parent process. This allows rollback and parent task tracking across agents.

An isolated sub-agent execution environment is used to perform tasks within a separate context, preventing state leakage or interference. It enforces reduced budgets and allowed-tools for safe orchestration and debugging.

Yes, you can enforce max-iterations and allowed-tools on a forked sub-agent. The skill defines a clear scope for forked execution, enforcing budget controls and separate contexts to prevent interference.

To track parent tasks across spawned sub-agents, the skill maintains parent task tracking and supports rollback. This ensures you can monitor the parent process while the sub-agent executes in its isolated context.

Use forked execution when you need safe experimentation, testing, or debugging without risking the parent process. It is ideal when you require reduced budgets, isolated contexts, and strict tool enforcement to prevent leakage.

Context leakage occurs when execution is not properly forked. This skill prevents leakage by enforcing separate contexts, ensuring the sub-agent operates in complete isolation from the parent process during orchestration.