hook-protocol

Git worktree-based task persistence stores task, context, result, and status data in a structured directory. This allows multi-agent workflows to recover execution state and resume operations immediately after unexpected crashes or restarts.

Maintaining an auditable history for multi-agent coordination is achieved by versioning task execution states and evidence in a git-backed hook directory. This enables structured tracking and cross-machine synchronization via standard git push and pull operations.

To start executing a persistent task, initialize the worker by reading the assigned hook from the .forge/hooks directory. The stored task and context schemas provide the necessary parameters to begin processing immediately without manual state reconstruction.

Multi-agent crash recovery does not require external dependencies. It uses a git worktree-based hook system to persist task and evidence schemas directly in the repository, ensuring durable coordination without external database infrastructure.

Persistent task tracking stores structured schemas including task.json, context.json, result.json, and status.json. These schemas capture the full execution lifecycle, enabling durable dispatch, execution, handoff, and recovery across distributed workers.