infiltr is an open-source command-line tool that uses reinforcement learning to perform adversarial red-teaming on large language model (LLM) systems. It helps AI security researchers and developers identify and address vulnerabilities such as prompt injection and other attacks.
infiltr is an Autonomous agents & workflows project. It focuses on testing and identifying vulnerabilities in LLM systems using adversarial red-teaming techniques. It is built as an open-source project for AI security researchers and developers. infiltr is open source under the MIT license. It ships for the command line.
It is developed by sushaan-k, and it first shipped in 2026. The project is developed in the open on GitHub with 13 commits in the last 90 days. Among its 5 catalogued features are red teaming, prompt injection testing, and reinforcement learning.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Same category — not a similarity match