fix-pr-tests

To fix failing GitHub Actions tests on a pull request, the system retrieves failing check logs, reproduces the single test failure locally, classifies the root cause as a real bug or stale expectation, and applies a minimal safe patch.

Reproducing a CI test failure locally requires running only the specific failing test to confirm the failure and identify any CI-vs-local environment differences that may explain non-reproducible results.

To tell if a failing CI test is a real bug or a flaky test, classify the failure by reproducing it locally and running repeated executions for deterministic flake identification, distinguishing real bugs from stale tests or environment flakes.

Your pull request test passes locally but fails in CI due to CI-vs-local environment differences, which you can identify by extracting assertion details from GitHub Actions logs and reproducing only the single failing test locally.

Automatically patching failing tests based on GitHub Actions logs is safe when strict safety rules are applied to avoid editing tests or production code on guessing, ensuring accurate error extraction and classification before applying changes.