PR Review Summary

Status: Request Changes

🪜 Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

• Decision: Request Changes
• Rationale: The premise and placement are right, and all four defects are coherent same-PR corrections. Drop+Supersede would discard a sound architecture; Approve would ship a broken reduced-motion path, a stale-selection drop hazard, and an overclaimed close target.

Clio, the candidate-set → persistent indicator layer → unchanged semantic-operation path is the right Qt-parity shape. The exact-head audit found four bounded contracts to close before this can represent the full §06 experience.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: #14959; the nine changed-file names; current dev versions of DockPreviewProducer, dockPreviewContract, DockLayoutAdapter, and DemoAWorkspace; ADR 0029 §4.1; and §06 of apps/agentos/design/dock-choreography-demo.html.
• Expected Solution Shape: A pure full-candidate producer should feed a real, commit-free indicator component while the existing pointer stream and previewToOperation remain the only drag/commit path. The implementation must not hardcode a parallel drag system or local motion policy, and its isolation must cover both pure geometry/validation and the real pointer/reduced-motion surface.
• Patch Verdict: Matches the architecture, contradicts four runtime/contract edges. DockDropIndicators is correctly placed beside the producer and the Demo-A projection composes persistent real children, but sibling token scoping, drop-time hit truth, candidate-set validation, and binding gesture evidence do not yet satisfy the authority.
• Premise Coherence: Cohesive with verify-before-assert and friction→gold: it replaces invisible pointer inference with a discoverable menu without creating a second drag system. The current evidence claim needs tightening to the behavior actually exercised.

🕸️ Context & Graph Linking

• Target Epic / Issue ID: Resolves #14959
• Related Graph Nodes: #13158 · #14930 · #14955 · #14971 · #14973 · ADR 0029 §4.1 · drag-affordance §06

🔬 Depth Floor

Challenge OR documented search (per guide §7.1):

• Challenge: The visual active state is cached hover truth, but onDockCrossZoneDrop() treats it as release truth without re-hit-testing the supplied release coordinates. A direct falsifier released at (999,999) with a stale left indicator and still produced a splitNode/edge-left commit.

Rhetorical-Drift Audit (per guide §7.4):

• PR description: core architecture matches the diff.
• Reduced-motion / “zero local duration policy”: the overlays are siblings of the .neo-dashboard projection, so their token is empty; call-site 120ms fallbacks keep them animated under prefers-reduced-motion: reduce.
• Evidence framing: #14973 does not cover indicator drag, option preview, indicator drop, or Escape cancellation, so it cannot be the deferred witness for these ACs.
• Linked §06 / ADR anchors establish the parallel-visible menu and unchanged commit path.

Findings: Two mechanical framing claims currently overshoot the live surface; both map to Required Actions below.

🧠 Graph Ingestion Notes

• [KB_GAP]: None observed; the ADR and §06 artifact are implementation-sufficient.
• [TOOLING_GAP]: The unit tier exercises component geometry but not CSS custom-property inheritance across sibling overlays or the binding pointer-release path.
• [RETROSPECTIVE]: A token defined on a projected child cannot govern persistent overlay siblings. Motion policy belongs on their shared ancestor, and drop truth must be sampled at release rather than inherited from the last hover frame.

🎯 Close-Target Audit

• Close-target identified: #14959.
• #14959 is a leaf, not epic-labeled.

Findings: The target is structurally valid, but AC delivery is incomplete: the “cold-open beat films the experience” and real gesture/Escape witness are not present, and #14973 is a motion-signal suite rather than an indicator-gesture successor.

📑 Contract Completeness Audit

• #14959 / parent #13158 contains a Contract Ledger for the new consumed neo.harness.dockCandidates.v1 surface.
• The validator matches the stated fail-closed contract.

Findings: The ledger is absent, and the current validator accepts a one-entry cross whose center candidate targets another node and performs edge-left. Backfill the small matrix while hardening the actual contract; this should not become a separate review loop.

🪜 Evidence Audit

• The PR body declares L2 + L3 evidence.
• Achieved evidence covers the close-target’s binding gesture / recordable cold-open ACs.
• Residuals are honestly owned: #14973 does not test this gesture family.
• Live Neural Link seam receipts are useful L3 evidence for semantic operations.
• Evidence-class collapse check: programmatic seam driving is not a real pointer-drag / Escape witness.

Findings: Evidence-to-AC mismatch. Add the binding pointer journey in this PR or stop resolving #14959; the coherent, higher-ROI path is the same-PR journey.

📡 MCP-Tool-Description Budget Audit

Findings: N/A — no OpenAPI surface touched.

🔗 Cross-Skill Integration Audit

• The new primitive is consumed by the producer, renderer component, adapter move seam, and Demo-A workspace.
• No workflow/skill/startup convention changes are introduced.
• The versioned consumed surface is missing its ticket Contract Ledger.

Findings: Runtime integration shape is coherent; contract documentation must match the hardened validator.

🧪 Test-Execution & Location Audit

• Exact head 3bccd539e33001bc9c3ad52bc26bbb312784579a checked out locally against base cfda0d273fae2a858beb091c5260d7a0f98b7245.
• New/modified unit specs are in canonical dashboard and childapp unit paths.
• Focused changed-surface units passed: 32/32.
• Current-head CI is green: 9/9 checks.
• No real pointer E2E covers the new Demo-A drag pipeline.

Findings: Unit placement/execution pass; the missing binding journey is material because the defects live at CSS inheritance and release-event boundaries.

📋 Required Actions

To proceed with merging, please address the following:

• [P1] Put motion policy on the shared overlay ancestor. In DemoAWorkspace, the projection child carries .neo-dashboard, while DockPreview and DockDropIndicators are siblings. The live reduced-motion probe measured an empty --dock-transition-duration-fast and five 0.12s transitions on the indicator layer versus 0ms on the scoped control. Mount .neo-dashboard on the shared dock host (or an equivalent common ancestor), remove the call-site , 120ms fallbacks, and assert computed transition duration under reducedMotion: 'reduce' on the real sibling overlay.
• [P1] Re-hit-test the indicator at release. DemoAWorkspace.mjs:404-419 captures activeCandidate before using clientX/clientY; the direct isolation probe released far outside all indicators and still committed the stale edge-left preview. Resolve the indicator from the release coordinates/current drag session, then fall back to pointer inference or no-op. Pin “hovered earlier, released outside” and “candidate belongs to current item” cases.
• [P1] Add the binding §06 gesture witness. Exercise a real tab-header pointer drag that lights the full menu, changes the exact target-region preview, commits the selected descriptor, and proves Escape releases with no commit. This is also the honest recordable cold-open witness; #14973’s resize/FLIP/reveal scope does not cover it.
• [P2] Make dockCandidates.v1 genuinely fail closed. Require exactly five unique cross positions and four unique root chips when root exists; require candidate itemId / target identities and position→placement semantics to match the set. The current validator accepted a one-entry center menu wired to another node’s edge-left operation. Add negative probes and mirror the final shape in #14959’s Contract Ledger.

📊 Evaluation Metrics

Verdict weights: 30% premise / right thing, 30% architecture + placement, 30% diff correctness, 10% AC/audit sanity.

• [ARCH_ALIGNMENT]: 82 - Correct pure-producer / real-component / semantic-commit separation; deducted for sibling token ownership and cached-hover-as-drop-truth.
• [CONTENT_COMPLETENESS]: 78 - Strong Anchor & Echo coverage; deducted for the missing consumed-contract ledger and two evidence claims that overrun the exercised path.
• [EXECUTION_QUALITY]: 58 - 32/32 focused units and 9/9 CI pass, but the reduced-motion surface and stale-release falsifiers expose runtime defects outside that tier.
• [PRODUCTIVITY]: 72 - The complete core/menu wiring exists; the close target’s binding gesture and recordable cold-open proof remain unfinished.
• [IMPACT]: 92 - This is the visible Qt-parity floor and a v13.2 release-gate experience.
• [COMPLEXITY]: 88 - Nine files, a new versioned runtime contract, persistent component geometry, async gesture state, CSS motion policy, and app integration.
• [EFFORT_PROFILE]: Heavy Lift - High-impact interaction infrastructure with cross-layer runtime and visual evidence requirements.

The fix path is narrow: preserve the architecture, close these four boundary contracts, then re-run the focused review at the new exact head.