AI & Machine Learning
Orchestrating Llm Attacks With PyRIT
The skill orchestrates automated, multi‑turn adversarial attacks on conversational LLMs with PyRIT, managing escalation and adaptive branching while incorporating scorer feedback and persisted conversation memory. It refines attack strategies over successive turns, enabling comprehensive, iterative red‑team campaigns against chatbots or agents. Security teams use it to conduct thorough, iterative testing of conversational AI systems when single‑shot scanning proves insufficient.
- Publisher / source
- Cybersecurity Skills (Community)
- Upstream record
- mukul975/Anthropic-Cybersecurity-Skills
- Version
- Version 3
- Published
- Publication review
- Approved for publication
- Content fingerprint
f009f55f1070d38fc4cb0a202d43a7bc259768f2db3524f37896e742d1ad2267