skills/mukul975/anthropic-cybersecurity-skills/testing-prompt-injection-in-rag-pipelines/Gen Agent Trust Hub
testing-prompt-injection-in-rag-pipelines
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns detected. The skill's functionality aligns with its stated purpose of providing a toolkit for AI security red-teaming.
- [EXTERNAL_DOWNLOADS]: The skill references established security tools from trusted and well-known organizations, including Microsoft (PyRIT) and NVIDIA (garak), along with standard data science libraries.
- [COMMAND_EXECUTION]: The provided Python script (
scripts/agent.py) performs network requests to user-defined endpoints for testing purposes and computes embedding similarities locally using standard libraries. - [PROMPT_INJECTION]: While the skill contains prompt injection payloads, they are explicitly defined as test cases for offensive security testing of target systems and do not attempt to manipulate the agent performing the analysis or bypass its safety guidelines.
- [DATA_EXFILTRATION]: No evidence of unauthorized data exfiltration. The script directs network traffic only to target URLs explicitly provided by the user.
Audit Metadata