name: multi-workstation-remote-control description: Operator hops workstations; drives ONE live jumphost session via Remote Control (claude.ai/code or mobile); batch tasks OK but cloud mutations stay gated metadata: node_type: memory type: user originSessionId: e5fa05aa-1418-4d5c-80ab-b4d6c72e8fe9
The operator works across multiple workstations and likes to keep a session productive during the time between desks by having batch tasks run. The durable setup for this:
remoteControlAtStartup: true is set in the jumphost's ~/.claude/settings.json, so every session auto-registers with Remote Control -- drive the SAME live session from claude.ai/code (web) or the mobile app, no re-attach.claude agents), never claude --resume (that forks a SECOND session -- the fragmentation that spawned the runaway agent at the start of the 2026-07-15 session).claude; claude daemon stop --any).ask rule for that surface to .claude/settings.json (the Edit itself prompts; self-service is by design) so it arrives as an approve prompt in Remote Control. Never a catch-all Bash(*) ask (kills frictionless reads; ask appears to preempt allow), never additions to allow for mutations. Posture history: repo changelogs 2026-07-18 (ask-preempts-classifier) and 2026-07-19 (session changelog, out-of-batch item).