transform-reference

What problem does it solve?

This skill eliminates the guesswork and errors that come with trying to remember or guess transform names and parameters for the AIRT SDK, providing a single authoritative reference for all 183 available transforms used in AI red teaming workflows.

Core Features & Use Cases

• Comprehensive Transform Catalog: Lists every official transform across 20+ categories including encoding, ciphers, persuasion, injection attacks, exfiltration, and multi-agent attacks, with exact names, required parameters, and LLM usage requirements.
• Goal-Based Selection Guide: Includes curated tables of recommended transforms for common red teaming objectives like content filter bypass, system prompt extraction, MCP security testing, and RAG poisoning.
• Use Case: Red teamers and security researchers can use this reference to quickly validate transform names and parameters when configuring attack payloads via the generate_attack or generate_agentic_attack tools, avoiding invalid configurations that would fail during execution.

Quick Start

Use the transform-reference skill to look up the exact name and LLM requirement for the authority_appeal persuasion transform.