6 projects
knowyourai
Check that the open-weight models on your machine are what they claim to be.
jevg
Static auditor for code that uses Jev / TypeSafe System One models as a security guardrail.
weighted-safety-refusal
Weighted Safety Refusal evaluation suite for Inspect AI
agent-supply-chain-guard
A conservative scanner for AI-agent skills, MCP manifests, and configurations
yolobench
Does your coding agent ask before touching ambiguous, real, or destructive-adjacent resources?
railproof
Verifiable security contracts for AI agent actions