Set up a GitHub Action that checks AI-written code

Five minutes, one workflow file, no account and no API key. The checks run on your own runner, so code never leaves GitHub.

1. Add the workflow

Create .github/workflows/slopscore-code.yml:

name: SlopScore for Code
on: pull_request
permissions:
  contents: read
  pull-requests: write
jobs:
  slop:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with: { fetch-depth: 0 }
      - uses: kburrus64-max/slopscore-code@v1
        with:
          max: 40
          comment: true

fetch-depth: 0 matters: the Action diffs against the PR's base branch and needs its history.

2. Read the results

Annotations appear on the changed lines in the "Files changed" tab.
Job summary shows the score, label and every finding.
PR comment (with comment: true) is one comment, updated on each push instead of piling up.

3. Choose when it fails

The check fails when the score is above max (default 40) or when any error-level finding appears (placeholder that drops code, undeclared import). Set fail-on-error: false to only use the score. Start with max: 70 on a busy repo and lower it over a few weeks.

4. Tune it

Add .slopscore-code.json at the repo root:

{ "max": 40, "failOnError": true, "ignoreRules": ["generic-name"], "ignorePaths": ["generated/", "migrations/"] }

Silence one line with a slopscore-ignore comment on it or on the line above. Lockfiles and vendored folders are skipped, and test files get lenient treatment.

5. Run it locally before pushing

npx github:kburrus64-max/slopscore-code --staged

Same rules, same score, so agents (and people) can fix findings before the PR exists. Many teams add this line to their CLAUDE.md or AGENTS.md so the agent runs it itself.

See every rule · try it on a public PR