Chain-of-thought leakage in AI agents: keep reasoning out of user replies
Chain-of-thought leakage is a production risk for AI agents. Learn where reasoning traces escape, why channel adapters matter, and how OpenClaw reduces exposure.
Tag archive
AI Agent Security
14 articles connected to this topic.
AI agent security boundaries in OpenClaw 2026.5.27
AI agent security boundaries in OpenClaw 2026.5.27 separate untrusted prompts, tool execution, network exposure and approvals so agent failures stay contained.
AI agent auth profiles: separate model logins without breaking agent runs
AI agent auth profiles separate model credentials by provider, runtime, and operator so self-hosted agents can migrate logins, recover cleanly, and reduce credential blast radius.
xAI device code OAuth for headless AI agents
xAI device code OAuth lets headless AI agents authorize from SSH, containers, and remote hosts without a localhost browser callback or pasted API key.
AI agent policy checks in OpenClaw 2026.5.20: what changed and how to use them
AI agent policy checks in OpenClaw 2026.5.20 add a practical control layer for channels, approvals, sandbox visibility, and workspace repair.
AI agent tool policies: sender-scoped permissions for dangerous tools
AI agent tool policies should vary by sender, channel, and action risk. OpenClaw 2026.5.12 adds sender-scoped controls for safer self-hosted agents.
How to sandbox AI agent code execution on a self-hosted setup
A practical 2026 guide to sandboxing AI agent code execution on your own hardware. Compares Docker, gVisor, Firecracker microVMs, and ephemeral containers, with a recommended setup for self-hosted agents.
Semantic Kernel RCE shows why AI agent tool inputs need real boundaries
Microsoft's Semantic Kernel RCE research shows how prompt injection becomes code execution when agents can influence tool parameters. Here's how to reduce the blast radius.
OpenClaw skill security checklist: what to review before you install
A practical checklist for reviewing OpenClaw skills and MCP servers before they get access to your files, accounts, shell, or APIs.
AI Agent Audit Logs: What to Capture Before Agents Take Action
AI agent audit logs need identity, authority, prompts, tool calls, policy checks, and outcomes. Use this checklist before autonomous workflows go live.
Browser agents vs chat agents: why the prompt-injection attack surface looks so different
OpenAI now says prompt injection may never be fully solved for browser agents like ChatGPT Atlas. Here is what that means for chat-channel agents and self-hosted setups.
How to Vet AI Agent Skills: A 6-Step Security Checklist
Vet AI agent skills before installing them with this 6-step security checklist: source trust, permissions, prompt injection, scripts, sandbox testing, and updates.
Arctic Wolf Launches the 'World's Largest Agentic SOC' — Hundreds of AI Agents Now Run Security Operations
Arctic Wolf's Aurora Agentic SOC deploys hundreds of specialized AI agents to handle security operations, powered by a 'Swarm of Experts' framework. It's agent-led, human-supervised — and available at no extra cost to existing customers.
Geordie AI Wins RSAC 2026 Innovation Sandbox — AI Agent Governance Takes the Crown
Geordie AI, a security and governance platform built specifically for AI agents, was named 'Most Innovative Startup' at RSAC 2026's Innovation Sandbox contest. The win signals that agent governance is now the hottest problem in cybersecurity.