ralph

Persistent execution loops for iterative task verification run by wrapping ultrawork inside a self-referential loop, reading instructions from `.agents/workflows/ralph.md`, and applying independent verifier checks to gate correctness across multiple rounds.

A self-referential execution loop for workflow automation is a persistent process that repeatedly coordinates task execution and verification across iterations, needed when refining deliverables or debugging agent behaviors where outputs must be validated across multiple rounds.

Independent verifier checks can gate task execution by running them inside a persistent self-referential loop that wraps ultrawork, validating results against intended behavior before proceeding to the next iteration.

Workflow-driven execution with verifier checks requires model invocation to be disabled, relying on a deterministic process defined in the repository workflow file to ensure consistent correctness gating across iterative automation scenarios.

Limitations of a self-referential loop for iterative automation include the requirement to disable model invocation and the need for a deterministic workflow file at `.agents/workflows/ralph.md` to drive execution, making it unsuitable for non-deterministic or ad-hoc tasks.

Debugging agent behaviors using an iterative execute-and-verify process involves running the agent inside a persistent loop that performs independent verification checks on outputs across multiple rounds, applying correctness gating until the intended behavior is achieved.