Explore Skills
Bounding and securing AI systems' tool and data access
Specifies least-privilege tool and data scopes for an AI agent or RAG system (including MCP servers and connectors), adds approval gates for consequential actions, and tests resistance to direct and indirect prompt injection.
1job relies on it most
1Path teaches it
0open now
Paths that teach it
Coming soonDesigning Controlled Automations and AI-assisted Workflows