factory-ralph-loop

Automate iterative lint fixing and test completion by running an agent in a loop that uses the filesystem as persistent memory, retaining code changes and progress until all success criteria are met.

Iterative PRD-driven work is automated by repeatedly invoking an agent with deterministic prompts and a simple loop orchestrator, using filesystem-backed state to preserve artifacts until tasks are exhausted.

An agent loop uses filesystem-backed state by writing progress and artifacts to the filesystem between iterations, ensuring the agent retains context and changes across repetitive runs without losing data.

Yes, a simple loop orchestrator controls agent iterations by repeatedly invoking the agent with deterministic prompts and checking explicit exit conditions like passing tests or clean lint results.

Explicit exit conditions are needed when an iterative agent loop must stop automatically upon meeting success criteria, such as when all tests pass, lint is clean, or PRD-driven tasks are fully exhausted.

The Ralph Loop operates without external dependencies, relying solely on the filesystem for persistent memory and a simple orchestrator to control automated iterative problem solving until exit conditions are met.