Blog

Agent security, in depth

Education and field notes on AI-agent security, capability-based policy, MCP enforcement, and the engineering behind eunox.

Blocking the lethal trifecta: the one attack only the MCP layer can stop

An agent with read-access to secrets, an exfiltration channel, and untrusted input is the canonical agentic-AI attack. Each tool call is individually authorized — the danger is the sequence, read_credentials → write_external, and eunox’s sequenceBlock condition is the only practical place to break that chain.

Read →

The prompt injection problem: why every AI agent needs a policy layer

Why fixing prompt injection inside the model doesn’t work, and why the only reliable defense is a policy layer at the structured tool call.

Read →