long-running-tasks

Long-running task resumption uses JSON-based state persistence files and a session-resume mechanism to recover multi-step workflows that exceed a single context window. It reloads prior progress to continue autonomous execution seamlessly across sessions.

Multi-step workflow progress tracking uses TodoWrite-style task management and JSON state files to enforce a single active feature. This ensures transparent, persistent state monitoring across multiple sessions for complex, autonomous jobs.

The Initializer and Coding Agent pattern splits complex refactoring tasks by separating initialization from iterative work. It documents resumption context, enabling autonomous agents to execute and persist multi-step migrations reliably.

Interrupted multi-session migrations recover without losing state through JSON-based workspace isolation and persistent progress files. The resumable task mechanism restores the exact workflow state, preventing data loss during unexpected interruptions.

Yes, multi-step job coordination supports both background and foreground subagents. It uses persistent state files to manage these subagents, enabling reliable recovery and resumption of resumable tasks after interruptions.

State persistence is needed for autonomous AI agent workflows that exceed a single context window or span sessions. It enables reliable resumption and progress tracking for complex refactoring and migrations that require multi-session execution.