Summary
agent-bundle dev on a project already owned by another dev process fails with:
[{"code":"AB5000","message":"Another agent-bundle dev process owns this project (pid 156163).","severity":"error"}]The pid is enough to kill the other process, but the common case is wanting to reuse it — and the message gives no way to find it. I had to dig the port out of ss -tlnp (users:(("node",pid=156163,…))) to discover the existing workbench at 127.0.0.1:48721.
Suggestion
The owner already persists enough state to enforce the lock; include its loopback URL (and maybe start time) in the AB5000 payload, e.g.
Another agent-bundle dev process owns this project (pid 156163, http://127.0.0.1:48721, started 41m ago).
That turns a dead end into a one-click reuse, especially for agents driving the workbench programmatically.
Observed on preview 560124af.
Summary
agent-bundle devon a project already owned by another dev process fails with:[{"code":"AB5000","message":"Another agent-bundle dev process owns this project (pid 156163).","severity":"error"}]The pid is enough to kill the other process, but the common case is wanting to reuse it — and the message gives no way to find it. I had to dig the port out of
ss -tlnp(users:(("node",pid=156163,…))) to discover the existing workbench at127.0.0.1:48721.Suggestion
The owner already persists enough state to enforce the lock; include its loopback URL (and maybe start time) in the AB5000 payload, e.g.
That turns a dead end into a one-click reuse, especially for agents driving the workbench programmatically.
Observed on preview
560124af.