successor-handoff

Mid-run agent handoff maintains context continuity by spawning successor agents when context pressure rises. It uses file-based memory and a lean parent orchestrator that reads only status files, ensuring long-running autonomous workflows continue seamlessly without human intervention.

File-first subagent management stores subagent contexts on disk rather than in memory. This approach keeps the parent orchestrator lean by reading only status files, enabling dynamic successor dispatch for workflows exceeding a single context window.

Set up autonomous agent handoffs by initializing workflow tracks for autonomous execution using the successor-handoff mechanism. Configure file-based memory and a lean parent orchestrator to dynamically spawn successor agents as context pressure rises during overnight research.

Yes, a parent orchestrator can manage subagent contexts without overflow by remaining lean and reading only status files from disk-based memory. This minimizes context usage while dynamically dispatching successor agents to handle the workload.

Dynamic successor dispatch is needed when an autonomous workflow exceeds a single Claude context window. It is ideal for multi-hour experiments or overnight research requiring a continuous agent presence without human intervention to reset context.

No, mid-run handoffs are designed for autonomous workflows operating without human presence. By automatically spawning successor agents when context pressure rises, the workflow maintains continuity for long-running jobs like overnight research independently.