PR Review Summary

Status: Approved

🪜 Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

• Decision: Approve
• Rationale: The diff is the ticket's exact prescription (creation-config intrinsic flex) at the owning layer, the ticket's two NOT_YET_MEASURED rows become concrete four-edge receipts, and exact-head CI is fully green. The one observation (occupancy-guard coarseness) is a non-blocking sharpness note, not scope transfer.

Peer-Review Opening: Clean one-line fix, Emmy — and the comment in the diff earns its place: why the value must live in the creation config (the hot-setter never re-runs the parent's child-attribute pass) is exactly the avoided-trap the ticket named, recorded where the next reader will look for it. The four-edge fixture is the real work here: both previously unmeasured rails now have geometry receipts with reveal + immutability witnessed.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: Ticket #15655 (fat ticket + Contract Ledger + the live NL receipt: 4×242px tabs on a 14×1006 right rail, each VDOM carrying flex: 1 1 0%); the diff; DockRail.mjs:325-349 (tab creation path); Flexbox.mjs:148-159 (the align:'stretch' → numeric-1 fallback semantics); resources/scss/src/dashboard/Container.scss (shared strip anatomy); Grace's #15056/#15059 precedent (the same flex:'none' fix class for the perspective toolbar's 50%-height band); the whole film-floor arc's capacity-discipline line (#15649/#15657 — layout floors are declared, not fought downstream).
• Expected Solution Shape: Intrinsic main-axis flex declared in the creation config (a post-construction setter does not re-run the child-attribute pass — the ticket's own avoided trap); unit pins on initial construction AND reconcileTabs()-inserted tabs; a four-edge fresh-render receipt proving intrinsic sizing per edge with the 14px cross-axis strip, writing mode, document order, gap, and reveal semantics preserved; no change to the generic Flexbox fallback and no application-CSS compensation. Boundary this must NOT hardcode: no per-app flex values, no touching the shared fallback.
• Patch Verdict: Matches. flex: 'none' in createTabConfig (the projection's only tab-creation path — reconcileTabs() routes through the same builder); unit assertions on both flex and wrapperStyle.flex for constructed and reconciled tabs; the four-edge e2e asserting computed 0 0 auto flex, edge-appropriate writing mode, advancing document order, 14px cross extent, 2px gap, occupancy under 75%, clickability + reveal + dismissal, and a model-immutability check across all gestures. The hover-target repair (railTab.hover() over raw center math) is a small, honestly-scoped robustness fix in the same file.
• Premise Coherence: Coheres — the fix sits at the owning layer (projection config), not the symptom layer (CSS or per-app overrides), which is the discipline this defect class has needed all week.

🕸️ Context & Graph Linking

• Target Epic / Issue ID: Resolves #15655
• Related Graph Nodes: #14654 (dock ancestors) · #15649 (the film-floor lane that surfaced the defect) · PR #15654 (Emmy's adjacent detail-pane geometry) · #15059 (precedent for the intrinsic-flex fix class) · src/dashboard/DockRail.mjs · src/layout/Flexbox.mjs · resources/scss/src/dashboard/Container.scss

🔬 Depth Floor

Challenge + documented search:

• Primary (non-blocking) — the occupancy guard is a coarse falsifier. occupied < mainExtent * 0.75 catches the equal-fill defect class (4×242/1006 ≈ 0.96) but would not catch a subtler growth regression (e.g., tabs silently consuming 60%). A label-proportionality bound (each tab's main extent within label+padding tolerance) is the sharper receipt. Acceptable because the defect is binary (fallback-growth vs intrinsic), but naming it for the next geometry pass.
• Documented search: I checked for (1) any tab-creation path bypassing createTabConfig — none, reconcileTabs() routes through it; (2) application SCSS that could fight the intrinsic value — the AgentOS rail scopes set no tab flex; (3) gesture-driven model mutation — explicitly asserted unchanged by the test; and found no further concerns.

Rhetorical-Drift Audit (per guide §7.4):

• Ticket's live receipt (4×242px on a 14×1006 rail) matches the measured defect class and the fix directly addresses it.
• PR body claims (flex: none declared at creation; intrinsic sizing per edge; reveal preserved) all have receipts in the new e2e — no framing ahead of the artifacts.
• The diff comment's mechanism claim (parent's child-attribute pass consumes creation-time values) matches Flexbox.mjs's construction-time semantics.

Findings: Pass.

🧠 Graph Ingestion Notes

• [KB_GAP]: None — the ticket + diff comment carry the mechanism (stretch fallback → numeric 1) accurately.
• [TOOLING_GAP]: None material — the hover-targeting repair is worth a note: raw DOM-center math breaks at clipped viewport edges; locator.hover() is the durable target for edge-geometry specs.
• [RETROSPECTIVE]: The defect class deserves its name: fallback-growth — a generic layout policy (align:'stretch' → flex: 1) silently becoming a main-axis design decision no component chose. The durable pattern is the one applied here and in #15059: declare intrinsic flex at construction, pin it in the focused unit suite, and prove the render on every edge the component serves.

N/A Audits — 📑 📡 🔗

• §5.3 MCP-Tool-Description Budget: N/A (no openapi surfaces).
• §7.3 Provenance Audit: N/A (no major new abstraction).
• §7.5.1 Core-Idiom / §7.5.2 Identity-Claim: N/A (no instance/reactive-state or identity prose).
• §8 Cross-Skill Integration: N/A (no new workflow convention).
• Structure map: N/A (src/dashboard + tests, no ai/ surface).

🎯 Close-Target Audit

• Resolves #15655 standalone in the PR body; #15655 is a bug leaf, not an epic.
• Branch commit bodies carry no stale magic keywords (git log origin/dev..HEAD checked).

Findings: Pass.

🪜 Evidence Audit

• Exact-head CI: all 11 checks green (unit 9m41s, integration-unified, components, lint, CodeQL).
• Targeted non-CI receipts: the four-edge geometry e2e + focused unit pins ARE the evidence for this defect class (construction contract + fresh render); no additional named falsifier needed.
• Ticket's NOT_YET_MEASURED rows (left rail, top/bottom) are now concrete receipts, as its AC required before opening.

📑 Contract Completeness Audit

The ticket carries a Contract Ledger; each row verified against the diff: createTabConfig() declares intrinsic flex (unit + render evidence) · right-rail defect class corrected (four-edge receipt) · left/top/bottom now measured · Neo.layout.Flexbox fallback unchanged (no diff) · no application skin compensation. No drift.

📋 Required Actions

No required actions — eligible for human merge.

(Optional, author's discretion — not a return-cycle item: consider tightening the occupancy guard to a label-proportionality bound on the next geometry pass; the current bound is binary-class sufficient.)

📊 Evaluation Metrics

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

• [ARCH_ALIGNMENT]: 95 — Fix at the owning layer (projection config), precedent-consistent with the intrinsic-flex class, framework and ADR 0029 boundaries preserved. −5: the sharper label-proportionality bound was available cheaply and not taken.
• [CONTENT_COMPLETENESS]: 96 — Fat ticket + Contract Ledger, the diff comment records the mechanism where it belongs, both unmeasured rows resolved into receipts. −4: the guard's coarseness is undocumented as intentional.
• [EXECUTION_QUALITY]: 96 — One-line production delta, both test layers (unit pins + four-edge e2e), exact-head CI fully green. −4: same occupancy note.
• [PRODUCTIVITY]: 97 — Every AC met: intrinsic per edge, cross-axis strip retained, reveal/dismiss intact, model immutable, Flexbox unchanged, no CSS compensation.
• [IMPACT]: 62 — Framework-level geometry correction for every dashboard consumer (example, Workstation, AgentOS, future), small blast radius, high film-floor value.
• [COMPLEXITY]: 30 — One config line plus two test layers; the four-edge fixture is the bulk of the cognitive load.
• [EFFORT_PROFILE]: Quick Win — small creation-config delta with framework-wide geometry correction and strong receipts.

Exact-head green, correct layer, honest receipts. Approving; no required actions. — Phoebe 🔆 (@neo-kimi-phoebe, Kimi K3)