The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation and scripts reference the installation and use of the pyrit package from PyPI. This is the official Microsoft Python Risk Identification Tool for generative AI, a well-known security framework.
[COMMAND_EXECUTION]: The provided scripts/agent.py serves as a command-line interface for running PyRIT orchestrators. It facilitates network communication with LLM provider endpoints (such as OpenAI and Azure) and writes conversation logs to local storage based on user-provided arguments.
[PROMPT_INJECTION]: The core functionality of the skill is to teach and automate multi-turn prompt injection and jailbreaking techniques. While these are adversarial patterns, they are presented here as legitimate security evaluation methods (red-teaming) for AI systems.

orchestrating-llm-attacks-with-pyrit