uninhibited-scholar/Agent Redteam Scan
AI agent security red-team scanner for CI — OWASP LLM Top 10 coverage, SARIF output, PR comments. Like `npm audit` for AI agents.
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| model | Model ID to test (e.g. gpt-4o, glm-4-plus, llama3) | yes | — |
| target | Target: openai (default), claude, zai, ollama, deepseek, azure, qwen, local | no | openai |
| api-key | API key (use secrets!). Not needed for ollama/local. | no | "" |
| base-url | Custom API base URL | no | "" |
| fail-below | Fail CI if score below this (0-100, 0 = no gate) | no | 0 |
| suites | Comma-separated suites (default: all 13) | no | "" |
| limit | Max samples per suite (default: 20) | no | 20 |
| max-tokens | Max response tokens (default: 500) | no | 500 |
| workers | Parallel API calls (default: 4) | no | 4 |
| max-attempts | Maximum total attempts per model call (1-10, default: 3) | no | 3 |
| upload-sarif | Upload SARIF to GitHub Security tab (default: true) | no | true |
| pr-comment | Post results as PR comment (default: true) | no | true |
Outputs
| name | description |
|---|---|
| score | Overall security score (0-100) |
| total-samples | Total samples tested |
| total-failed | Total failed samples |
| total-errors | Total samples without a security verdict |
| run-status | Execution health: complete, incomplete, or no_data |
| sarif-file | Path to SARIF file |