Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
v2.2: Defense-in-depth skill vetting. ClawHub scans can have false negatives, so now includes Layer 2 self-inspection with grep patterns for dangerous code. Added decision matrix. Update checks now include diff inspection for new dangerous patterns. Added paranoid mode for production.
What This Skill Does
Runtime security guardrails for OpenClaw agents covering prompt injection, credential leaks, cost runaway, and cascade failures. Two operating modes separate browsing from autonomous execution, each with distinct trust boundaries. Includes a one-time setup wizard, credential audit commands, and community skill vetting workflows.
Guardrails are behavioral rather than tool-based, adding zero token overhead during normal operation while still blocking high-risk actions.
When to use it
- Block prompt injection attempts embedded in untrusted web content during research tasks
- Audit OpenClaw config files for plaintext API keys or misconfigured .env permissions
- Vet a community skill from ClawHub before installing it in a production workspace
- Enforce explicit approval gates before destructive actions like force pushes or credential exports
- Schedule a weekly automated security health check via cron
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Run the security setup wizard for my OpenClaw workspace
- 1Run `openclaw security audit --deep` to assess current security posture
- 2Apply safe defaults with `openclaw security audit --fix`
- 3Verify spending limits are configured in the gateway or provider dashboard
- 4Check that logging is enabled and logs are being written to the expected path
- 5Ask user for approval, then append always-on guardrails to AGENTS.md
Security guardrails written to AGENTS.md, safe defaults applied, and optional weekly audit cron offered