ralph-loop

An autonomous task loop keeps Claude executing multi-step processes until a defined completion state is met. It prevents premature exits by using a stop hook to monitor progress and re-inject prompts, ensuring complex workflows are fully resolved without interruption.

Automating folder processing involves starting an iterative loop that handles each item, generates plans, and moves processed files to a Done directory. The loop continues until all items are resolved, outputting a specific completion token when finished.

File system state changes trigger task completion by monitoring for file movements to a designated directory like Done. This flexible completion strategy stops the autonomous loop when the observed file state matches the defined success condition.

Yes, a configured stop hook is required to run autonomous loops. The hook monitors task progress during execution and re-injects prompts as needed, ensuring the agent continues iterating until the explicit completion state is reached.

Promise-based completion stops an autonomous loop by detecting a specific output token like TASK_COMPLETE. When the agent outputs this token, the loop recognizes the multi-step process is finished and safely terminates the iteration.

Workflows exit prematurely without an autonomous iteration loop to sustain execution. Implementing a continuous task loop with a stop hook prevents early termination, ensuring Claude processes all items until the final completion state is met.