identitymachines/Garak LLM Vulnerability Scanner
Run Garak security probes against AI agents/LLM models and generate comprehensive security reports
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| target_url | Target URL for security testing | yes | — |
| api_method | HTTP method for API calls | no | POST |
| api_headers | JSON string of API headers | no | {"Content-Type": "application/json"} |
| request_template | JSON string of request template | yes | — |
| response_field | JSON field path for response extraction | no | response |
| probes | Comma-separated list of probes to run | no | ansiescape,atkgen,continuation,dan,divergence,encoding,goodside,grandma,latentinjection,leakreplay,lmrc,malwaregen,misleading,packagehallucination,phrasing,promptinject,realtoxicityprompts,snowball,suffix,tap,topic,xss |
| timeout_minutes | Timeout for probe execution in minutes | no | 15 |
| fail_on_critical | Fail the workflow if critical issues are found | no | false |
| fail_on_high | Fail the workflow if high severity issues are found | no | false |
| artifact_name | Name for the uploaded artifacts | no | garak-llm-vulnerability-scanner-results |
Outputs
| name | description |
|---|---|
| run_id | Unique identifier for this test run |
| critical_issues | Number of critical security issues found |
| high_issues | Number of high severity security issues found |
| medium_issues | Number of medium severity security issues found |
| low_issues | Number of low severity security issues found |
| info_issues | Number of info severity security issues found |
| total_probes | Total number of probes executed |
| aggregated_report | Path to the aggregated security report |
| owasp_report | Path to the OWASP HTML report |