Run loopd goals in OTHER projects from one manager session. Standing goals per project keep developing; manager watches and steers.
- Target project needs the plugin loaded (its own instance drives its goals: engine + control-worker + schedule-worker are per-directory, started lazily on first event/goal).
- Manager registers targets in
.opencode/loopd/managed.json(absolute paths) or via envLOOPD_MANAGED_DIRS(colon-separated, like PATH). - Create cross-dir goal:
loopd_create_goal({..., directory: "/abs/target"}). State, artifacts, checks, worker session live in target. Turn 1 is prompted directly bystart()— no manager engine needed.
- List across projects:
loopd_status_all({})→ per-dir totals + blocked items withdirectoryfield. - Inspect foreign goal:
inspect_background_goal({goal_id, directory: "/abs/target"})orlist_background_goals({directory}). Omitdirectoryfor current project. - Steer:
send_goal_input/nudge_goalby goal ID (engines route via global session IDs; notifications reach manager). - TUI: header shows managed count + store path;
:projectslists managed dirs with counts.
- Checks run from target root (
checkCwddefaults there). Fences are per-dir stores — no cross-talk. - Scope inference already takes
directory. - Goal IDs are UUIDs — always display directory alongside goal.
- Target dir needs its own
opencode.jsonc(model, permissions). Manager cannot grant what target denies. - Standing goals:
scheduleEveryMswithoutscheduleMaxRuns= infinite dev loop;interactive:truewhere human-gated. - Managed staff never self-complete into blocked on a foreign store: cross-dir
creation defaults
interactive:trueunless you passinteractive:falseexplicitly. Every turn then comes from manager steering, so a transient fence (foreign shell running, quiescence gate) parks instead of burning turns into evaluator rejections andblocked. - Fenced workers park, never spin: a
Workspace is busy — held by: …denial names the exact holder (operation, ticket, uncontrolled call, legacy write with session, goal, files, age). Callawait_quiescence, report progress, end the turn. Release paths (afterWorkspaceTool, operation end) fire parked awaits with inbox evidence; the engine's idle continuation picks the worker up. Owner-only tools (force_*,loopd_release_workspace, nudge, send, pause/resume/clear) fail closed for worker sessions — never call them from inside a goal.
P1 investigation showed goalService.start(directory,...) keys everything off the passed dir, and v1 session.create query.directory / v2 root fallback location.directory pin the worker in target. Per-directory plugin instances (one per open project) own the 30s/5s/1s timers for their dir. A manager trio would double-drive. So P2 = verified not-needed; aggregation (P3/P4) is the whole feature.