Blog
Running AI agents without losing sleep.
What goes wrong when agents get write access, and the controls that stop it. Built from real incidents.
Claude Code
Claude Code --dangerously-skip-permissions: how to use it safelyWhat Claude Code's --dangerously-skip-permissions flag (YOLO mode) skips, what still runs, and five ways to keep auto mode without deleted folders or databases.
OWASP
OWASP Top 10 for Agentic Applications (2026), explained with controlsThe OWASP Top 10 for Agentic Applications (ASI01 to ASI10) in plain language, with real incidents and the controls that address each risk.
Prompt injection
Prompt injection in AI agents: how to stop data exfiltrationIndirect prompt injection turns AI agents into data leaks. Real incidents (GitHub MCP, Supabase MCP, EchoLeak), the lethal trifecta, and controls that work.
Guardrails
AI agent guardrails: input, output and action guardrails comparedWhat AI agent guardrails are, the three kinds (input, output, action), what each catches, and benchmark numbers on approval fatigue and false positives.
SOC 2
SOC 2 for AI agents: the change-management evidence auditors will ask forWhen AI agents change production, SOC 2 CC8.1 still applies. What auditors will ask about agent changes, and how to produce approval and audit-trail evidence.
Incident
How an AI agent deleted a prod database and how to stop itReplit's agent wiped a production database during a code freeze. What happened and the controls that would have stopped it.
Approvals
Human-in-the-loop approval for AI agents: what to ask aboutHow to design approvals that catch the risky 1% without the fatigue.
MCP
MCP security checklist: 10 controls for AI agent toolsTokens, untrusted content, config tampering, malicious servers, approvals and logging.
EU AI Act
EU AI Act Article 14: human oversight for AI agentsWhat Article 14 asks for, and how to produce the evidence for agents that act.
Comparison
Squidbrake vs Kontext: AI agent control comparedHuman approval, open source, MCP, self-hosting and price.