PR Review Summary

Status: Request Changes

🪜 Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

• Decision: Request Changes
• Rationale: #16412 is valid and the derive-don’t-assert direction is right, so this is repair-in-place rather than Drop+Supersede. The current derivation duplicates only part of the downstream stable-topology predicate and still emits a false privileged hint for valid layout changes.

Peer-Review Opening: Deriving the tour-reset flag, failing closed on malformed documents, and repairing the old invalid fixture are all the right direction. The blocker is one boundary mismatch: Workspace currently proves less than DockFlip and the reconciler require from geometryOnly.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: #16412; exact changed-file list and head e4082b2723a20790731bfd099d79255b2f303f71; current Workspace, DockTopologyDiff, DockProjectionReconciler, DockFlip, and their focused specs; ADR-0029; the ticket Contract Ledger; hosted CI and author receipts.
• Expected Solution Shape: A tour reset may pass geometryOnly:true only when the before/after documents preserve the exact topology/projection identity the downstream in-place path assumes: node identity, type, ancestry, order, orientation, active/item projection, and auto-hide state; size-only deltas remain admissible. Prefer one authority for this predicate or thread the reconciler’s actual staged/in-place outcome, rather than duplicating a weaker category test in Workspace.
• Patch Verdict: Contradicts the expected boundary. Workspace.isTopologyStableReset() checks errors plus four item-diff arrays, while exact execution admits valid orientation, child-order, active-tab, and auto-hide changes that the production reconciler rejects from its stable path or projects into a different tab/rail structure.
• Premise Coherence: Coheres with verify-before-assert in replacing a literal with a derivation; conflicts at implementation depth because the new proof is not the proof the consumer relies on.

🕸️ Context & Graph Linking

• Target Epic / Issue ID: Resolves #16412
• Related Graph Nodes: ADR-0029, #16391, PR #16403, #14650, DockProjectionReconciler, DockFlip.play({geometryOnly})
• Origin Session ID: e034ddc4-234b-4d72-8858-80780abf4527

🔬 Depth Floor

Challenge: The implementation assumes an empty moves/adds/removes/tabReorders projection proves the full downstream stable-topology contract. Exact valid-document counterexamples falsify that assumption: the helper returns true for orientation-only, child-order-only, active-tab-only, and edge-auto-hide-only changes.

Rhetorical-Drift Audit (per guide §7.4):

• PR description: drift — “Topology is the four structural categories” is narrower than the production reconciler’s admitted invariant, and excluding autoHideFlips is not safe for edge-item tab/rail projection.
• Anchor & Echo summaries: the new method’s intent is documented and free of temporary ticket/line anchors.
• [RETROSPECTIVE] tag: N/A — none added.
• Linked anchors: #16412/#16403 establish that geometryOnly is a consumer-trusted proof obligation.

Findings: Drift is behavioral, not prose-only: the narrower definition is the implementation defect named below.

🧠 Graph Ingestion Notes

• [KB_GAP]: The ticket’s four-array shorthand under-describes the existing stable-topology contract; downstream source additionally binds orientation, child ordering, active/item projection, and auto-hide projection.
• [TOOLING_GAP]: The current Workspace controls omit four valid-document negative axes, so the partial predicate can stay green.
• [RETROSPECTIVE]: A trusted optimization flag is a proof obligation. Deriving it from a parallel, weaker summary is still an assertion; share the downstream authority or carry the downstream outcome.

N/A Audits — 📡 🔗

N/A across listed dimensions: this two-file Workstation behavior fix changes neither MCP/OpenAPI descriptions nor skills, conventions, or cross-substrate primitives.

🎯 Close-Target Audit

• Close-targets identified: #16412
• #16412 confirmed not epic-labeled — it is an open bug / ai / refactoring leaf.

Findings: The target is valid, but not yet resolved: the false geometryOnly declaration remains reachable on valid resets, and AC4’s recorded-tour receipt remains unchecked.

📑 Contract Completeness Audit

• #16412 contains a Contract Ledger for Workspace.refreshDockWorkspace → DockFlip.play.
• The PR matches it exactly: the chosen computation does not prove the consumer’s “no topology swap can be pending” contract for all valid document axes.

Findings: Contract drift flagged at the producer→consumer proof boundary.

🪜 Evidence Audit

• PR body contains an Evidence: declaration line.
• Achieved evidence covers the close target: the L3 claim is unit evidence with refreshDockWorkspace stubbed; it does not witness the reconciler/DockFlip interaction or the ticket’s recorded-tour reset after structural divergence.
• Residuals are truthful: the body says Residual: none, while both Post-Merge Validation items remain unchecked and AC4 explicitly asks for the recorded tour behavior.
• Two-ceiling distinction is described in Post-Merge Validation, but not reflected in the greppable evidence line or close-target disposition.
• No external receipt is incorrectly used as a current-head merge gate.

Findings: Evidence/AC mismatch. After the behavioral repair, capture the routed runtime receipt or preserve it as an explicit residual and keep #16412 non-closing.

📜 Source-of-Authority Audit

• Authority checked: ADR-0029 plus the production DockProjectionReconciler stable-path checks and DockFlip.play({geometryOnly}) contract.
• Finding: DockTopologyDiff category emptiness is a useful mutation summary, not the complete admission authority for the in-place projection path. Workspace must not locally redefine the stronger downstream invariant.

🧪 Test-Evidence & Location Audit

• Execution evidence: all 15 hosted checks green at exact head e4082b2723; author focused receipts are current-head-appropriate.
• Reviewer falsifier: executed the actual Workspace.isTopologyStableReset() over valid documents. Orientation-only, reversed split children, active-tab-only, and edge-auto-hide-only deltas all returned true; the production reconciler rejects or reprojects each corresponding delta.
• Test location: the added Workspace.spec.mjs placement is correct; the behavioral matrix is incomplete.

Findings: Falsifier failed the patch. Green CI covers the implemented four-array predicate, not the downstream contract.

📋 Required Actions

To proceed with merging, please address the following:

• Make Workspace prove the same invariant its downstream geometry-only consumers require, rather than only checking that four DockTopologyDiff item arrays are empty. At exact head e4082b2723, the actual helper returns true for valid active-tab-only, edge-auto-hide-only, split-orientation-only, and structural child-order-only resets; DockProjectionReconciler rejects those deltas from its stable path while Workspace still forwards geometryOnly:true into DockFlip.play. Derive admission from an exact document-topology projection covering node identity/type/ancestry/order/orientation plus active/item/auto-hide projection state while still admitting size-only deltas, or thread the reconciler’s actual in-place/staged outcome to DockFlip so the declaration cannot survive a staged fallback. Add exact tests for those four counterexamples and retain unchanged, resize-only, and malformed controls. Do not use computeShapeFingerprint() alone because it omits node IDs and can collide when same-shaped children reorder. After repair, capture AC4’s recorded-tour reset receipt; if the available seat cannot reach it, declare that residual and keep #16412 non-closing instead of claiming Residual: none.

📊 Evaluation Metrics

Verdict weights: 30% premise / right thing, 30% architecture + placement, 30% diff correctness, 10% AC/audit sanity. These are importance-to-verdict weights, not effort budgets.

• [ARCH_ALIGNMENT]: 70 - correct derive-don’t-assert direction and placement, but the producer duplicates a weaker predicate than the consumer’s authority.
• [CONTENT_COMPLETENESS]: 72 - strong JSDoc and malformed-input treatment; the declared topology definition and no-residual claim omit real axes and AC4.
• [EXECUTION_QUALITY]: 52 - focused and hosted suites are green, yet four valid-document counterexamples recreate the false declaration.
• [PRODUCTIVITY]: 78 - compact two-file change with useful fixture repair; one coherent boundary correction remains.
• [IMPACT]: 90 - geometryOnly gates a topology-swap bypass on the flagship Workstation path.
• [COMPLEXITY]: 72 - topology identity spans document structure, active projection, auto-hide rails, and an asynchronous reconcile/flip boundary.
• [EFFORT_PROFILE]: Heavy Lift - bounded surface but a high-integrity cross-class invariant and runtime receipt.

The derivation seam is the right place to fix this; it just needs to derive the whole consumer contract.

[review-budget-managed]

• outcome: within-budget
• ordinary-limit: 2
• activation-issue: 15257
• activation-pr: 15307
• activated-at: 2026-07-16T20:54:31Z