An AI reviewer
for the code and content you ship.

# Score what HEAD did to a file vs HEAD~1
$ judge score-edit src/auth.ts --judge code-quality

→ scoring before (HEAD~1)…  72.4
→ scoring after  (HEAD)  …  68.9

✗ verdict: REGRESSED   weighted Δ −3.5
  ↓ input_validation  0.9 → 0.6  (−0.3)
  ↓ error_handling    0.8 → 0.5  (−0.3)
  = type_safety       1.0 → 1.0   (·)

history: 12 runs → 71.6 → 71.8 → 72.4 → 68.9
scorecard: judge.tools/p/8a7…f2c (cited evidence)
exit 2  (--fail-on-regression)

The wedge: edit scoring. Diffs HEAD~1 → HEAD, scores both, returns a verdict with per-metric deltas. Plug it into a pre-commit hook with --fail-on-regression and you stop shipping silent quality drops.

# Score a PR diff through a rubric
$ judge score --judge code-quality \
$   --pr acme/api#482 \
$   --target-kind REPO --target-key acme/api

→ fetch PR diff via GitHub
→ submit_score (forced tool-use)

● overall 74.0 / 100   △ +1.2 IMPROVED
  ✓ type_safety      0.90   (+0.10)
  ✓ error_handling   0.75   (+0.15)
  ✗ test_signal      0.40   (·)

rationale (cited):
  login.ts:42 — new guard validates token
  no tests added for the new branch

The GitHub PR source fetches the diff and a rubric scores it — the exact run the GitHub App does on every PR. Overall in 0–100, per-metric deltas vs the target's last score, and a cited rationale you can audit.

# Once: scaffold .judge/config.json + GH Action
$ judge init && judge install gh-actions

# On every PR (CI):
$ judge run --fail-on-regression

→ syncing 3 pipelines from .judge/config.json
✓ pr-diff:code     78.2  IMPROVED  (+2.1)
✓ auth:security    74.0  STABLE    (·)
✗ readme:doc       61.5  REGRESSED (−4.3)

1 regression → exit 1, build fails
PR comment posted with 3 scorecards

Check in .judge/config.json with named pipelines (PR diff, a file, README — anything text). One judge run in CI scores them all, fails the build on any REGRESSED, and the GitHub App posts the scorecards to the PR.