skills/mukul975/anthropic-cybersecurity-skills/orchestrating-llm-attacks-with-pyrit/Gen Agent Trust Hub
orchestrating-llm-attacks-with-pyrit
Pass
Audited by Gen Agent Trust Hub on Aug 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads and installs the Microsoft PyRIT library from the official Python Package Index (PyPI) as part of its setup process.
- [COMMAND_EXECUTION]: Provides a functional Python script (
scripts/agent.py) designed to orchestrate complex, multi-turn adversarial conversations with target AI models based on user-provided objectives. - [DATA_EXPOSURE]: Includes instructions for managing OpenAI API credentials via environment variables or
.envfiles, which is a standard and recommended practice for development tools.
Audit Metadata