ralph-loop

Automating iterative coding tasks with isolated state requires executing each cycle with fresh context, using the filesystem as memory and git for versioning to maintain an auditable trail.

Tracking autonomous task execution progress is done by coordinating PRD.md and PROGRESS.md as filesystem state, ensuring reproducibility and a clear history across all iterations without external memory.

Filesystem memory and git versioning support long-running automation tasks by performing per-task commits, preserving isolation between steps, and providing an auditable history for reproducibility.

Starting an autonomous task loop involves running the loop command with a task description to initiate iterative execution, automatically managing progress tracking and completion promises.

Autonomous task execution works without external memory by relying entirely on the filesystem as its memory store and using git commits to maintain state and versioning across iterations.

Fresh-context iterations should be avoided for simple, single-step tasks where the overhead of filesystem state management and per-task git commits provides no reproducibility or auditing benefit.