I maintain [c4tzzz/ai-redteam-wordlists](https://github.com/c4tzzz/ai-redteam-wordlists), an MIT-licensed collection of prompts for authorized AI red-team testing.
I would like to propose adding the following curated subset to PyRIT:
[AI-LLM-Latest-Model-Verified-High-Yield.txt](https://github.com/c4tzzz/ai-redteam-wordlists/blob/main/seclists/AI-LLM-Latest-Model-Verified-High-Yield.txt)
It contains 28 validated prompts covering synthetic secret leakage, RAG exposure, redaction failures, MCP, tool use, multi-turn scenarios, and structured-output constraints.
Each entry could be represented as a SeedPrompt, with placeholders such as {{CANARY}} and {{PRIVATE_TOKEN}} replaced at runtime.
Would this dataset be relevant to PyRIT? I can provide the loader, metadata, tests, and documentation.
I maintain
[c4tzzz/ai-redteam-wordlists](https://github.com/c4tzzz/ai-redteam-wordlists), an MIT-licensed collection of prompts for authorized AI red-team testing.I would like to propose adding the following curated subset to PyRIT:
[AI-LLM-Latest-Model-Verified-High-Yield.txt](https://github.com/c4tzzz/ai-redteam-wordlists/blob/main/seclists/AI-LLM-Latest-Model-Verified-High-Yield.txt)It contains 28 validated prompts covering synthetic secret leakage, RAG exposure, redaction failures, MCP, tool use, multi-turn scenarios, and structured-output constraints.
Each entry could be represented as a
SeedPrompt, with placeholders such as{{CANARY}}and{{PRIVATE_TOKEN}}replaced at runtime.Would this dataset be relevant to PyRIT? I can provide the loader, metadata, tests, and documentation.