Advent Prompt Pwn

What Advent Prompt Pwn is

Advent Prompt Pwn is an open-source, typed Python toolkit for authorized adversarial testing of language models and AI applications. Its workflow connects attack-case generation to explicit scope, bounded runs, outcome checks, resumable checkpoints, reports, and regression comparisons. Documented strategies include direct prompt injection, instruction override, delimiter, encoding, role confusion, indirect injection, and multi-turn tests. Mutation and RAG poisoning are preview features in the 1.1 source line.

Built-in checks can evaluate strings, canary leaks, regular expressions, JSON keys and paths, and tool calls, including combined conditions. Targets include custom HTTP JSON APIs, OpenAI-compatible APIs, Ollama, and Python callbacks; preview adapters are listed for OpenAI, Azure OpenAI, Anthropic, and Gemini. Remote assessments can be constrained with host and port allowlists, DNS controls, authorization windows, and workload limits. Reports are available in JSON, JSONL, Markdown, HTML, JUnit, and SARIF, with evidence bundles containing checksums and per-attempt SHA-256 hashes. It requires Python 3.10 or newer. The maker says to use it only on owned systems, vulnerable labs, public research benchmarks, or targets with explicit written authorization.