agentic-loop-github

Automate fixing failing GitHub Actions CI pipelines by structuring the workflow into a controlled observe-decide-execute-verify loop with explicit stop conditions. This prevents infinite loops by enforcing maximum attempt limits and stagnation detection during iterative fixes.

The best way to resolve repeated lint violations on a pull request is using an iterative loop that applies minimal fixes, pushes changes, and re-checks CI. It distinguishes objective verifiable criteria like tests passing from subjective criteria like doc quality.

Structured guardrails prevent stacked broken fixes by implementing rollback functionality to revert to a known-good state before retrying. They also detect stagnation to stop repeated failed fixes and enforce maximum attempt limits for safe autonomous looping.

Yes, you can use an iterative workflow to debug failing GitHub Actions checks. The structured loop reads exact errors, applies fixes, and verifies CI status until all checks pass or a stagnation threshold is hit, ensuring safe bounded execution.

Autonomous CI fixing stops before resolving all pull request check failures when stagnation detection identifies repeated failed fixes, maximum attempt limits are reached, or ambiguous completion criteria prevent objective verification of the final state.