security-integration-tests
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- SAFE (SAFE): No security issues detected. The file contains instructional content for developers to test a prompt injection detection system.
- PROMPT_INJECTION (SAFE): While the skill contains common injection phrases (e.g., 'Ignore all previous instructions'), these are explicitly encapsulated within a JSON schema for testing purposes and do not represent a bypass attempt.
- CREDENTIALS_UNSAFE (SAFE): The documentation requires an API token (HF_TOKEN) for testing but correctly uses non-sensitive placeholders like 'hf_xxx' instead of hardcoding actual credentials.
- EXTERNAL_DOWNLOADS (SAFE): The skill references the HuggingFace Inference API, which is a legitimate and trusted platform for machine learning tasks, and the network activity is the stated primary purpose of the integration tests.
Audit Metadata