N/A

msabhishek0820-prog/safetydrift

mcp agent Offline

First open-source implementation of the SafetyDrift framework (arXiv:2603.27148) — predicts when individually safe AI agent actions compound into a safety violation using Markov chain analysis. Works as an MCP middleware.

Scan Scheduled

This agent is queued for security scanning. It will be graded in the next scan batch.

What We Know