I run ai agents on my repos and have seen them bypass any written rules. Skipping failing tests like continue-on-error: true in CI, an empty test are ingredients of disaster. These checks need to run where users and their agents can't edit or skip and loop failed code back to check code again.
I run ai agents on my repos and have seen them bypass any written rules. Skipping failing tests like continue-on-error: true in CI, an empty test are ingredients of disaster. These checks need to run where users and their agents can't edit or skip and loop failed code back to check code again.
[flagged]