The Lint That Would Have Caught It Is Off by Default
Every Check Was Green. Five Guarantees Were Not.
A Scanner Flagged Our Supply-Chain Package. It Was Right.
The Switch Inside the Room
Govern Your Coding Agent in One Command
What Your Coding Agent Lets You Control — and Three Things We Got Wrong
The Prompt You Never Saw
Our Benchmark Caught Us Twice
Agents Need Their Own Computer. That's Half the Problem.
Your Agent Harness Is Already a Security Kernel
Governed Is Not Confined
Your Permission List Can't See Taint
We Skipped Permissions in Our Own Governance Tool
We Found a False Positive in Our Own Flagship Demo
Your Agent Just Learned to Write Programs. Can You Still Govern It?
My Own Sandbox Blocked Me. The Fix Wasn’t a Hole.
Does Taint Cross the Subagent Boundary? An Experiment
Why "Deny" is Dangerous: The Case for Absent Tools in AI
AI Aikido: Neutralizing Prompt Injection with Determinism
The ZombieAgent Threat: Why Your AI's Memory is a Ticking Time Bomb
Running Claude Code Safely: A Sandbox Setup Guide