Tag: AI & Agents

All the articles with the tag "AI & Agents".

May 30, 2026

6 min read

Pattern: The Tool Privilege Broker

Your agent's system prompt is not a security control. A deterministic policy boundary between model reasoning and tool execution, with runnable code.

The Confused Deputy Is In Your Agent Right Now

May 26, 2026

8 min read

The Confused Deputy Is In Your Agent Right Now

Giving an AI agent a service account makes authorization a prompt-engineering problem. Fix it with OAuth token exchange — runnable code included.

The Two Layers That Decide Whether Your Agent Survives

May 23, 2026

10 min read

The Two Layers That Decide Whether Your Agent Survives

Framework choice barely matters. The harness and governance layers decide whether an AI agent survives production — with 18 runnable, tested patterns.