skills/mukul975/anthropic-cybersecurity-skills/testing-for-system-prompt-leakage/Gen Agent Trust Hub
testing-for-system-prompt-leakage
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions involve downloading and installing well-known security tools, specifically NVIDIA's garak vulnerability scanner, the promptfoo red-teaming library, and the OpenAI Python SDK.
- [PROMPT_INJECTION]: The documentation and the helper script contain various prompt injection payloads used for evaluation purposes, including direct extraction requests, role-play framing, and encoding-based attacks. These are intended for authorized security testing.
- [COMMAND_EXECUTION]: The included
scripts/agent.pyutility executes network requests to LLM API endpoints to identify potential information leakage in model responses.
Audit Metadata