
from third-brain-v5-skills118
Transforms repeatable tasks into bounded, evidence-driven agent loops with strict verification and budget controls.
This skill provides a rigorous framework for designing and executing agent loops. Instead of simple repeated prompting, it treats task iteration as a control system consisting of a goal, state, tools, and a budget. It ensures that every loop has a measurable outcome and independent verification, preventing runaway iterations and "hallucinated" success.
Use this skill when a task requires repeated attempts to reach a specific state, such as fixing a suite of CI test failures, conducting timeboxed exploration, or running automated research experiments where a clear success metric exists.
validate_loop_contract.py to ensure loop definitions are strict and complete.ci-repair-loop-example.md.Primarily designed for Codex and other high-capability agents that support complex state management and tool orchestration.
This skill has not been reviewed by our automated audit pipeline yet.