Render harness dock split/tab layouts

Context

#13030 is the docking strategy scoping anchor under Epic #13012. Its first contract leaves are now merged: #13061 / PR #13062 defined the dock-zone model, #13086 / PR #13087 defined the transient dockPreview state, and #13089 / PR #13090 defined the split/tab adapter boundary.

All open PRs were verified empty at the start of this nightshift cycle. Ada has taken #13104 for the preview-renderer leaf. This ticket creates the next non-overlapping one-PR leaf: turn the accepted split/tab adapter contract into concrete renderable Neo layout/tab structures, without bundling preview rendering or persistence.

Release classification: post-release Agent Harness product-line leaf; Project 13, not the v13 release board.

The Problem

The docking model and adapter boundary are documented, but there is not yet a concrete implementation that turns dock-zone split and tabs nodes into renderable Neo structures. Without this leaf, the next implementation could accidentally re-open the already-settled contract questions or bundle too many concerns:

• split rendering versus preview-zone rendering;
• tab-slot rendering versus persistence;
• stable dock item identity versus runtime component references;
• existing hbox / vbox / tab.Container primitives versus a premature core Dock primitive.

The missing step is narrow: consume the model and adapter contract into existing Neo primitives so committed dock layouts can render as nested split containers and tabbed slots.

The Architectural Reality

V-B-A before filing:

• Live issue state at 2026-06-13T16:23:03Z: #13030 remains open as the scoping anchor; #13104 is open and assigned to Ada for preview rendering; no open split/tab rendering implementation ticket exists in the latest-open queue.
• Scope-specific GitHub sweep for split/tab rendering, split tab rendering, split/tab adapter, and docking renderable structures found only #13030, #13104, and the closed contract leaves #13089 / #13086 / #13061.
• Local exact sweep over resources/content/issues, resources/content/discussions, and learn/agentos/HarnessDockZoneModel.md found only the parent/contract references, including the explicit statement that concrete split/tab rendering is a follow-up leaf.
• A2A last-30 sweep at 2026-06-13T16:23:03Z showed the recent #13104 lane claim and prior lifecycle broadcasts, but no competing [lane-claim] or [lane-intent] for split/tab layout rendering.
• Knowledge Base semantic check found adjacent historical primitives (#3519 tab drag/drop, #2192 splitter, #8155 dashboard detach/reintegration), but no existing issue for implementing this exact dock-zone-to-split/tab rendering leaf.
• Labels were verified live: enhancement, ai, and architecture exist.

Source anchors already established by the merged contract:

• learn/agentos/HarnessDockZoneModel.md defines split projection as ordinary containers with layout: {ntype: 'hbox'|'vbox', align: 'stretch'} and child ordering/flex derived from the model.
• The same contract defines tabs projection as Neo.tab.Container-compatible config with item order and activeIndex derived from model-owned item ids.
• Runtime componentRef resolution is a handoff concern; serializable model identity remains stable even when live refs are stale or missing.

The Fix

Implement the split/tab rendering adapter that consumes the accepted dock-zone model shape and produces concrete renderable Neo structures.

Expected shape:

1. Read the current HarnessDockZoneModel.md contract from dev and use it as the source of authority for split and tabs node projection.
2. Add a narrow adapter surface owned by the dashboard/harness docking layer unless structural pre-flight proves a different owner is necessary.
3. Map split.orientation: 'horizontal' to hbox and split.orientation: 'vertical' to vbox, preserving child order and honoring valid ratio/flex data when present.
4. Map tabs.items and activeItemId into tab.Container-compatible configs while preserving stable dock item identity outside display text.
5. Resolve componentRef through a runtime resolver when available; if no live component or blueprint can be resolved, render a recoverable placeholder/fail-safe without corrupting the model.
6. Keep preview rendering (#13104) and layout persistence as separate consumers.

Contract Ledger Matrix

Decision Record Impact

Aligned-with ADR 0020 and Epic #13012. This ticket implements a harness/dashboard docking leaf and should not amend an ADR unless implementation evidence proves the adapter must become a general core layout primitive.

Acceptance Criteria

• A concrete split/tab rendering adapter exists for dock-zone model split and tabs nodes.
• Horizontal and vertical split nodes project to existing Neo layout primitives without introducing a parallel layout engine.
• Split child order is model-owned and preserved; valid size/ratio data is honored or normalized without mutating the source model.
• Tab nodes project to existing tab/container semantics with item order and active item behavior covered by tests.
• Runtime componentRef resolution is separated from serializable model identity; stale refs produce recoverable output instead of deleting or corrupting model items.
• The adapter has no dependency on transient dockPreview, DOMRect, pointer coordinates, or window geometry.
• Unit or static tests cover split projection, tab projection, missing/stale component refs, and boundary rejection of preview-only input.
• If implementation introduces or relocates any .mjs file, the PR documents structural-pre-flight evidence for the chosen destination.

Out of Scope

• Preview zone rendering during drag (#13104).
• Layout save/restore persistence.
• Cross-window choreography (#13025 / #13028).
• Partner-specific presets or public client naming.
• Neural Link mutation-to-commit write-through.
• Generalizing the adapter into a core Dock primitive without implementation evidence.

Avoided Traps

• Parallel layout engine - rejected. Existing Neo layout and tab primitives are the first implementation target.
• Preview coupling - rejected. This leaf renders committed layout state, not hover intent.
• Persistence bundling - rejected. Save/restore needs its own leaf after renderable structures exist.
• Core primitive too early - rejected. A dashboard/harness adapter can prove the shape before any core lift.

Related

Parent: #13012 Scoping source: #13030 Depends on: #13061 / PR #13062 and #13089 / PR #13090 Adjacent: #13086 / PR #13087, #13104, #13063, #13025, #13028

Retrieval Hint: "harness docking split tab rendering adapter dock-zone model hbox vbox tab Container componentRef placeholder"