yashchimata/judgegate

Statistical trust gate for LLM judges. Fails CI when the judge does not agree with human labels well enough to gate.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Stale
License
None
namedescriptionrequireddefault
labelsPath to the human-labeled examples file.yes
configPath to the judge.yaml configuration.yes
min-kappaTrust threshold override.no""
alphaSignificance level override.no""
no-probesSkip the probe battery.nofalse
commentPost the report as a sticky pull request comment.notrue
fail-on-inconclusiveFail the check when there are too few labels to decide.nofalse
github-tokenToken used to post the pull request comment.no${{ github.token }}
python-versionPython version used to run judgegate.no3.12
namedescription
verdictTRUSTED, UNTRUSTED, INCONCLUSIVE, or ERROR.
exit-codeRaw judgegate exit code (0, 1, 2, or 3).