OpenClaw Device Lab is a disposable, isolated, Codex-driven lab for investigating OpenClaw device identity, operator scope, pending approval, and stale request behavior outside Navid's real Second Brain and Nava setup.
GitHub repository: https://github.com/NavidBroumandfar/openclaw-device-lab
This repository is an autonomous lab workspace for planning, reproducing, testing, documenting, and potentially fixing OpenClaw device lifecycle issues in a controlled environment.
The lab uses:
- Lab OpenClaw profile: oc-device-lab
- Lab gateway port: 19791
- Agent identity: Device Lab Agent
- Local folder: /Users/navidbr/Projects/openclaw-device-lab
This lab is not the real Second Brain workspace, not the real Nava runtime, and not a production OpenClaw environment. It must not use real tokens, real device approvals, real operator state, real runtime services, or real autostart configuration.
OpenClaw device identity and approval behavior can involve profile state, gateway identity, operator scope, stale requests, and recovery paths. Those areas are risky to test against a real personal automation setup.
This lab exists so experiments can be planned and executed in a disposable boundary before any real-system action is considered.
The lab protects the real Second Brain and Nava setup by requiring:
- A dedicated local folder.
- A dedicated lab profile named oc-device-lab.
- A dedicated lab gateway port, 19791.
- Explicit forbidden boundaries for real profiles, real state, real tokens, real logs, and real services.
- Approval gates before public posting or any real-system operation.
- Validation before each autonomous commit and push.
Active autonomous lab.
EXP-1 through EXP-6 have established the disposable gateway baseline, constrained unsafe broad status surfaces, confirmed direct TCP/HTTP reachability probes, and confirmed a WebSocket challenge-only pre-connect probe. Signed connect, pending approval, operator scope upgrade, and stale request reproduction remain gated on a disposable identity and observation plan.