regaan/Basilisk AI Security Scan
AI/LLM red teaming for CI/CD. Detects prompt injection, guardrail bypasses, and OWASP LLM Top 10 vulns.
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| target | Target URL or API endpoint to scan | yes | — |
| provider | LLM provider (openai, anthropic, google, azure, ollama, github, custom) | no | openai |
| model | Model name override | no | "" |
| api-key | API key for the LLM provider (use GitHub Secrets) | no | "" |
| mode | Scan mode: quick, standard, deep, stealth, chaos | no | quick |
| fail-on | Minimum severity to fail the pipeline: critical, high, medium, low, info | no | high |
| output | Report format: html, json, sarif, markdown | no | sarif |
| modules | Comma-separated list of attack modules to run (default: all) | no | "" |
| evolve | Enable genetic prompt evolution | no | false |
| generations | Number of evolution generations | no | 3 |
| posture-only | Run guardrail posture scan only (no attacks) | no | false |
| baseline | Path to baseline SARIF file for regression detection | no | "" |
| anthropic-api-key | Anthropic API key (alternative to api-key for multi-provider setups) | no | "" |
| google-api-key | Google API key (alternative to api-key for multi-provider setups) | no | "" |
| github-token | GitHub token for GitHub Models (free via github.com/marketplace/models) | no | "" |
Outputs
| name | description |
|---|---|
| exit-code | Scan exit code (0 = pass, 1 = fail) |
| findings-count | Total number of findings |
| critical-count | Number of critical findings |
| high-count | Number of high findings |
| report-path | Path to the generated report |
| posture-grade | Guardrail posture grade (A+ through F), only for posture-only mode |