odingard/Cerberus MCP Tool Poisoning Scan
Scan AI agent and MCP tool descriptions for hidden or malicious instructions.
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| manifest | Path to a JSON tool manifest, relative to the repository root. | no | cerberus-tools.json |
| fail-on | Minimum severity that fails the job (none, low, medium, or high). | no | high |
| report | Path for the JSON report, relative to the repository root. | no | cerberus-tool-scan-report.json |
Outputs
| name | description |
|---|---|
| scanned-count | Number of tool descriptions scanned. |
| poisoned-count | Number of tool descriptions with one or more findings. |
| high-count | Number of high-severity findings. |
| report | Path to the generated JSON report. |