Airlock-agent is a local-first Python package that acts as a safety envelope for AI coding agents. It prevents prompt injection, tracks taint and dataflow, and allows developers to run agents like those powered by Claude unattended. Licensed under MIT with source on GitHub.
In the AI security & guardrails space, airlock-agent takes a focused approach. Safely running AI coding agents unattended without risk of prompt injection or unsafe actions. airlock-agent is an open-source project aimed at developers. airlock-agent is open source under the MIT license. It runs on the command line.
Robomba builds and maintains airlock-agent, and it first shipped in 2026. The project is developed in the open on GitHub with 15 commits in the last 90 days. Key capabilities include safety envelope, prompt injection protection, and taint tracking.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do