Checklist
OpenClaw config audit checklist
This is the fast version of what I look for in an OpenClaw audit. If you cannot answer these checks confidently, your setup probably has risk hiding in plain sight.
Secrets and exposure
- +API keys are not inline in openclaw.json or copied into public repos
- +Sensitive values load from a secrets provider or environment
- +Sensitive tool output is redacted in logs
- +Gateway is bound intentionally, not exposed by accident
Permissions and blast radius
- +Destructive actions are gated behind approval
- +DM and group access are scoped intentionally
- +Agent roles are separated by capability
- +Hooks or policies enforce rules that survive prompt drift
Spend and runaway protection
- +maxConcurrent is set for agents and subagents
- +Compaction thresholds are configured
- +Cheap models handle heartbeats and low-value work
- +Long-running tasks have tracking instead of disappearing into the void
Recovery and continuity
- +Sessions recover cleanly after restart or compaction
- +Important events flush to persistent memory
- +Operational files like AGENTS.md and HEARTBEAT.md exist and are useful
- +There is a known path for async work to push completion back
Want me to run this on your actual setup?
Email your config and I will turn this checklist into a written audit with findings, priorities, and specific fixes.
Get your free OpenClaw audit