Security Check
Tool Profile: Limit What Your Agent Can Do
OpenClaw agents have access to a broad set of tools by default. Tool profiles like "messaging" restrict this to a safe subset. Deny lists can block specific dangerous tools.
Why It Matters
The principle of least privilege applies to AI agents too. An agent that only needs to send messages shouldn't have filesystem access, shell execution, and browser control. Over-permissioned agents are higher-risk targets.
How to Fix
Set tools.profile to "messaging" (or another restrictive profile) as a baseline. Use tools.deny to block specific tools, and tools.allow to add back only what's needed.
openclaw.json
{
"tools": {
"profile": "messaging",
"deny": ["system.run", "browser.action"]
}
}