odingard/Cerberus AI-Agent Runtime Check
Reproducible Cerberus allow/block self-check for CI: a benign path runs, a dangerous one is blocked.
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| core-version | Version of @cerberus-ai/core to install (npm dist-tag or exact version). | no | 3.1.0 |
| node-version | Node.js version to run the check under. | no | 24 |
| fail-on | When to fail the job. "never" (default) is report-only — the check never breaks your build. "unexpected" fails the job only if protection did not behave as expected (a dangerous path was NOT blocked, or a benign path was falsely blocked). | no | never |
| report-path | Path (relative to the workspace) for the JSON evidence report. | no | cerberus-report.json |
| upload-artifact | Upload the JSON report as a workflow artifact. | no | true |
Outputs
| name | description |
|---|---|
| verdict | Overall verdict: PASS, FLAG, or FAIL. |
| report-path | Path to the JSON evidence report. |