Lift dock tear-out lifecycle with FleetCockpit as first consumer

Context

Epic #17539 lifted the shared reducer/projection loop into Neo.dashboard.DockWorkspace, then migrated Workstation and the showcase examples. Its reviewed closeout matrix still names O-3 as a blocker: the tear-out/window-lifecycle half has no engine owner and FleetCockpit—the zero-grant host selected as the first consumer—still extends plain Neo.container.Base.

The dependency is explicit rather than inferred. Epic #17335’s formal review stops its FleetCockpit decomposition until this O-3 leaf lands, then requires a fresh census; its native subgraph is still empty.

At origin/dev 2e1a63cf5c, apps/agentos/view/fleet/cockpit/Container.mjs is 4,085 physical lines / 1,795 code LOC. It repeats the class-owned dock holder loop (applyDockZoneOperation, getDockZoneDocument, onDockZoneDocumentChange, projectDockModel, refreshDockWorkspace, onDockCrossZoneDrop) and keeps the common tear-out mechanics beside product-specific popup and Fleet data behavior.

The full structure map currently fails before output with Cannot create a string longer than 0x1fffffe8 characters. Scoped maps succeeded: src/dashboard contains 30 files with DockWorkspace.mjs at 305 code LOC; the Fleet cockpit folder contains two files, with Container.mjs at 1,795 code LOC. No new file or directory is proposed.

The Problem

The base class owns the in-window reducer and projection transaction, but it stops immediately before the second repeated half: detach placement capture, admitted vessel ownership, exact semantic return, and worker window connect/disconnect routing. Workstation, Demo B, and FleetCockpit each implement those mechanics again.

The duplication is already behaviorally expensive. FleetCockpit cannot adopt the engine holder loop without threading its tear-out state through the existing hooks, so the flagship FM consumer remains on the retired host pattern. Conversely, lifting the whole app methods would be wrong: those methods also contain product policy that the engine must not know—Workstation/Demo B owner grants and staged embodiment, Demo B cross-window targets, and Fleet click-popouts plus owner-held return panes.

The reviewed epic boundary is therefore load-bearing: the engine owns admission, document mutation, and window lifecycle; the app owns embodiment and grant policy.

The Architectural Reality

A current-head three-host body comparison confirms the boundary rather than merely repeating the epic:

• All three capture {tabsNodeId, index} before detachItem, delete refused captures, and reintegrate through semantic addTab with exact-home/fallback behavior.
• All three maintain pre-terminal connections versus post-commit tearOutPanes, then reparent the same live pane into the admitted vessel.
• Workstation and Demo B consume owner grants, staged embodiment, native routes, and workspace-target registration; FleetCockpit deliberately has zero grant references and uses tearOutPaneHandles / returningTearOutPanes.
• Each top-level window handler also owns unrelated product flows. Those branches cannot move into the engine under a generic name.

DockWorkspace already provides the required holder hooks: getDockProjectionOptions, getPreservedItemIds, beforeRefreshDockWorkspace, resolvePane, and the serialized refresh transaction. The missing addition is a generic tear-out lifecycle with explicit hooks at the policy and live-instance boundaries, not another app helper or a generic DragZone change.

The Fix

1. Extend Neo.dashboard.DockWorkspace with the common tear-out state and lifecycle: placement capture/refusal cleanup, admitted connection versus committed ownership, semantic exact-position reintegration, and worker connect/disconnect registration for vesselFlow=tear-out.
2. Expose narrow template hooks for app-owned policy and embodiment:
   • optional grant admission/consumption (default no-op for Fleet);
   • live pane resolution/reparenting;
   • observable adopt and return moments;
   • product-owned openTearOutVessel / closeTearOutVessel;
   • non-tear-out window connect/disconnect continuation.
3. Keep generic lifecycle defaults inert unless a consumer enables dock tear-out through getDockProjectionOptions.
4. Migrate FleetCockpit to extend DockWorkspace as the first consumer:
   • inherit the holder/projection/cross-zone loop;
   • configure shell index 1 and flex: 1;
   • map syncControlBar, pane resolution, and preserved detached/torn ids onto existing hooks;
   • retain Fleet click-popout, data, UI, and same-instance returningTearOutPanes policy app-side.
5. Remove only the duplicated imports, fields, and methods whose ownership moved. Workstation and Demo B migrations onto the new tear-out half remain separate reversible leaves.

Contract Ledger Matrix

Decision Record impact

aligned-with ADR 0029 and the reviewed O-3 boundary on Epic #17539. No persisted dockZone.v1, wire vocabulary, or service topology changes.

Acceptance Criteria

• DockWorkspace owns the common tear-out placement/admission/return/window-lifecycle path without importing any app module or naming a product grant format.
• The engine path covers terminal-first and connect-first admission, refused detach, wrong host/flow/admission token, pre-terminal disconnect, committed disconnect, and exact-once semantic reintegration.
• Grant policy, vessel opening/closing, live-pane embodiment, and non-tear-out window branches remain explicit consumer hooks with inert defaults.
• Adopt and return lifecycle moments are observable, and FleetCockpit proves #16415's same-live-instance return through returningTearOutPanes.
• FleetCockpit extends DockWorkspace; the six duplicated holder/projection methods and app-owned preview producer are removed rather than shadowing the engine implementation.
• Fleet maps its persistent toolbar, shell index, pane resolver, control-bar sync, and detached/torn preservation onto existing DockWorkspace hooks with no Fleet-only hook added to the engine.
• The rendered root keeps both neo-dock-workspace and fm-fleet-cockpit; the projected .neo-dashboard token floor and current light/dark Fleet values remain unchanged.
• Existing click-popout, gesture tear-out, splitter resize, perspective switch, cross-zone drop, and DockService/NL journeys stay green; the tear-out witness crosses the real gesture threshold.
• A source mutation that bypasses the engine lifecycle or shadows an inherited holder method turns the focused matrix red.
• Workstation and Demo B remain behaviorally unchanged and keep their app implementations until their own migration leaves.

Out of Scope

• Migrating Workstation or Demo B onto the new tear-out lifecycle.
• Cross-window workspace transfer, vessel conversion/parking, click-popout product flows, or a new grant schema.
• Fleet information design, data feeds, SCSS redesign, or pane decomposition.
• Changing DragZone.proxyParentId, DockTearOut gesture grammar, persisted documents, or neo.harness.* wire identifiers.
• Closing Epic #17539; this leaf discharges O-3 only.

Avoided Traps

• Core-only Fleet parent swap. Leaves O-3's tear-out/window lifecycle owner unresolved and certifies the wrong blocker.
• Moving whole app window handlers into the engine. Imports product flows and grant policy into a generic class.
• One hook per Fleet sequence. Hook admission requires a named engine lifecycle moment, an inert minimal-consumer default, and product policy left app-side.
• Recreating panes on return. Violates the owner-held live-instance law and loses in-flight state.
• Migrating all three rich hosts in one PR. Hides behavior changes and destroys independent rollback.

Related

Parent: #17539

Blocking predecessor for: #17335

Related: #17541 Related: #17546 Related: #17630 Related: #16415 Related: #16322

Live latest-open sweep: checked the latest 20 open issues at 2026-08-24T07:00:57.607Z; exact all-state and local-history searches found no equivalent. A2A in-flight sweep: latest 30 messages across read states at 2026-08-24T07:00:57.607Z; no competing claim. Semantic ticket sweep found no equivalent.

Origin Session ID: 0dc1379e-5329-4fba-80ca-f6466822f7c9

Retrieval Hint: FleetCockpit DockWorkspace O-3 tear-out lifecycle engine admission document mutation window connect adopt return hooks same live pane