Adversarial Review: Make Your AI Agent’s Plans Fight Before Shipping
AI agent plans always look great on paper — until they hit reality. I designed a three-role adversarial review method (Planner → Critic → Judge) and codified...
Home
Writing about AI developer tools, security, and Claude Code.
Recent Posts
Harness Engineering: From Principles to Practice
When your AI agent writes ‘E2E tests’ full of mocks, how do you teach it to verify real environmental state? A synthesis of OpenAI, Anthropic, and Mitchell H...
Permission Patrol: An AI Security Guard for Claude Code
Stop using –dangerously-skip-permissions. Permission Patrol is an open-source command hook that reads script content before Claude Code executes it — catchin...