INTRODUCING AUGUSTUS: OPEN SOURCE LLM PROMPT INJECTION TOOL

Summary from the TLDR InfoSec newsletter

Augustus is an open-source Go-based LLM vulnerability scanner that tests models against 210+ adversarial attacks across 47 categories, including jailbreaks, prompt injection, encoding exploits, data extraction, and agent manipulation, shipping as a single binary with support for 28 LLM providers. Inspired by NVIDIA's garak but reimplemented in Go for faster execution and zero runtime dependencies, Augustus features a pipeline architecture with composable probes, buff transformations (encoding, paraphrase, poetry, and low-resource language translation), and 90+ detectors, including LLM-as-a-judge and HarmJudge. The tool is part of Praetorian's "The 12 Caesars" campaign.