data-sciencetech/ForceField AI Security Scanner
Scan for prompt injection, PII leaks, and LLM attacks. Run security evals with 116 built-in attack prompts.
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| mode | Scan mode: selftest, audit, eval, or both | no | both |
| sensitivity | Detection sensitivity: low, medium, high, critical | no | medium |
| audit-path | Path to scan for hardcoded prompts and PII (used in audit mode) | no | src/ |
| install-extras | pip extras to install (e.g. ml, all) | no | ml |
| fail-on-detection | Fail the workflow if selftest detection rate drops below threshold | no | true |
| detection-threshold | Minimum detection rate (0-100) to pass. Only used if fail-on-detection is true. | no | 95 |
| eval-suite | Path to a custom eval suite YAML file (used in eval mode) | no | "" |
| eval-categories | Comma-separated attack categories for built-in eval (used in eval mode without eval-suite) | no | "" |
| python-version | Python version to use | no | 3.12 |
Outputs
| name | description |
|---|---|
| detection-rate | Selftest detection rate (e.g. 100) |
| detected | Number of attacks detected |
| total | Total number of attacks tested |
| audit-issues | Number of audit issues found |
| eval-passed | Number of eval cases that passed |
| eval-failed | Number of eval cases that failed |
| eval-detection-rate | Eval detection rate (0-100) |