Detects prompt injection by its effect on a sacrificial canary model, not just pattern matching: untrusted input hits a powerless model first, a behavioral check reads the…
Tag
#defense-in-depth
MCP servers, skills and plugins tagged “defense-in-depth” on GitHub.
1 entries · real GitHub data