diff --git a/.ciagent/CHECKPOINT.json b/.ciagent/CHECKPOINT.json
index 278c0c3..44900f6 100644
--- a/.ciagent/CHECKPOINT.json
+++ b/.ciagent/CHECKPOINT.json
@@ -1,8 +1,8 @@
 {
-  "phase": 3,
-  "stage": "complete",
+  "phase": 4,
+  "stage": "verify",
   "milestone": "v0.1",
-  "phase_role": "execution",
+  "phase_role": "final",
   "attempts": 0,
-  "updated_at": "2026-09-27T02:10:00Z"
+  "updated_at": "2026-09-27T02:30:00Z"
 }
\ No newline at end of file
diff --git a/.ciagent/PLAN.md b/.ciagent/PLAN.md
index 3b9c693..14185c9 100644
--- a/.ciagent/PLAN.md
+++ b/.ciagent/PLAN.md
@@ -1,64 +1,50 @@
-# PLAN — Milestone v0.1 Phase 3: v2 Surface Mockups + Polish + 3-Click Validation
+# PLAN — Milestone v0.1 Phase 4 (FINAL): Review + Owner Approval Gate + Ship
 
 ---
-phase: 3
+phase: 4
 milestone: v0.1
-status: planned
+status: executed
 must_haves:
-  - "Project graph mockup panel: cross-linked sources and findings across projects, rendered from mock data (REQ-60)"
-  - "Task inbox mockup panel: open questions become checkable follow-ups, rendered from mock data (REQ-60)"
-  - "Typography care pass: reader experience per vision §4 — report body measure/leading, section rhythm, small-caps/labels audit across all screens (REQ-61)"
-  - "≤3-click happy path validated on defaults: Ask → Choose → Run initiated in 3 clicks, verified by test (REQ-61)"
+  - "Full-milestone review: PASS-WITH-FIXES verdict — all 5 P0s fixed and verified (drawer separators, empty-question guard, border-hairline token, fixture URL honesty, README accuracy)"
+  - "Audit: PASS — reconstruction test, file discipline, branch hygiene, commit discipline, tag verification, critical-issue check all green"
+  - "Final phase patch v0.1.4 shipped (pending owner approval for milestone merge to main)"
 personas:
-  - frontend-engineer
-requirements: [REQ-60, REQ-61]
+  - ci-lead-developer
+requirements: [REQ-62]
 ---
 
 ## User-Facing Surface
 
-- **`/library`** — the v2-surface mockup page: two panels rendered from fixture data — (1) **project graph**: nodes for research topics/projects with cross-links to shared sources and findings (visual mockup: SVG or CSS positioned nodes with labeled edges, clearly marked "v2 preview"); (2) **task inbox**: open questions extracted from past runs as checkable follow-up items with status toggles. A page-level banner explains these are v2 vision mockups, not yet functional.
-- All existing screens get the typography care pass: consistent section rhythm, reading measure enforced on long-form text, caption/label treatment unified.
-- Navigation affordance: the Reading Room and run screens gain a "Library (v2 preview)" link so the mockup page is discoverable in the demo.
+The prototype deliverable presented for owner approval: the full 5-screen click-path (Ask → Choose → Plan & Run → Reading Room → Library preview), README quickstart, releases v0.1.0–v0.1.3, and this phase's review-driven polish (P0 fixes) as v0.1.4-pending.
 
-## Happy Path
+## Happy Path (owner review script)
 
-1. Owner runs the full P1+P2 path (ask → choose → approve → trace → report)
-2. From the report screen, clicks "Library (v2 preview)" → sees project graph with 4–6 topic nodes sharing source edges + task inbox with 3–5 checkable items
-3. Toggles a task's checkbox → item visibly flips status (client state only)
-4. Back on /ask: with defaults (no context, Deep Report pre-recommended), the owner reaches a streaming run in exactly 3 clicks: Research this → (arrive /choose) Start research → (arrive /run) Approve plan & start research
-5. The 3-click criterion is pinned by an automated test, not just manual belief
+1. `cd frontend && pnpm install && pnpm dev` → http://localhost:3000
+2. Ask → question → Research this (click 1)
+3. Choose → Deep Report recommended card → Start research (click 2)
+4. Plan & Run → Approve plan (click 3) → live trace → pause/resume/cancel
+5. Read the report → hover [2] → exact quote; ledger panel; export Markdown/JSON
+6. Library (v2 preview) → graph + inbox
+7. Owner decision: approve → v0.1.4 + milestone merge to main; or request changes
+
+## What this phase did
+
+1. **Review** (multi-persona, full milestone diff `main..HEAD`): verdict PASS-WITH-FIXES — 5 P0s (all owner-visible) + 2 P1s + P2 notes. All P0s fixed in-phase: dials separator rendering (P1-1 flag from P1), empty-question guard (P1-2 flag from P1), border-hairline Tailwind token (21 surfaces were near-black), fixture URL switched from a real arXiv link (resolving to an unrelated paper) to example.org, README screens table rewritten for all 5 routes. P1s: report title hardcodes Deep Report shape (documented prototype limitation), getRun fallback lacks a demo-run caption (documented). P2 notes logged for v0.2 hardening.
+2. **Audit**: AUDIT PASS — tags v0.1.0–v0.1.3 verified annotated + on milestone branch; REQ-52..61 spot-checked as genuinely delivered; secrets never committed; branch hygiene clean; commit discipline PASS (2 zero-loss deviations noted); gates re-run green 42/42.
+3. **Ship (this commit sequence)**: PR #4 (phase/04 → milestone), squash-merge, tag v0.1.4, release with full milestone summary. **Milestone merge to main happens ONLY after recorded owner approval (REQ-62)** — the pipeline halts at the gate.
 
 ## UX Acceptance Criteria
 
-- Graph: nodes labeled with topic names; edges show shared-source counts; hover on node highlights it (title attribute acceptable); visually distinct from production screens ("v2 preview" banner + muted styling)
-- Inbox: items show question text + originating run status; checkboxes toggle open↔done; at least one item pre-checked
-- Typography: body text ≥16px, reading measure ≤68ch on report/ask/choose long-form, consistent heading scale, no clipped or truncated labels
-- 3-click test: automated test counts discrete user clicks from /ask to a streaming run using default methodology — must be ≤3
-- Quality gates: `pnpm typecheck && pnpm lint && pnpm test && pnpm build` green
-
-## Tasks (wave-ordered; persona: frontend-engineer unless noted)
-
-### Wave 1 — Fixtures
-- **T-3.1**: Library fixtures in mock layer: `getProjectGraph()` (nodes: topic/project + linked sources + findings) and `getTaskInbox()` (items: question, source run, status) — extend ApiClient interface (both map cleanly to v0.2 GET /api/graph + GET /api/tasks) (REQ-60)
-
-### Wave 2 — Library page
-- **T-3.2**: `/library` route with v2-preview banner; ProjectGraph panel (SVG/CSS nodes + labeled edges from fixtures); TaskInbox panel with toggleable checkboxes (client state) (REQ-60)
-- **T-3.3**: Navigation links: Reading Room + run screens gain "Library (v2 preview)" link; root nav hint if layout has one (REQ-60)
-
-### Wave 3 — Typography pass
-- **T-3.4**: Audit + fix typography across screens: reading measure on long-form, section rhythm (spacing-section usage), caption/label letter-spacing consistency, heading scale order (display → headline → body) (REQ-61)
-
-### Wave 4 — 3-click validation
-- **T-3.5**: Automated 3-click test: /ask → Research this (1) → /choose → Start research on recommended default (2) → /run → Approve plan (3) → assert trace streaming; document the count in the test name (REQ-61)
-
-### Wave 5 — Gate (persona: ci-lead-developer)
-- **T-3.6**: Boundary review + gate run; P1-flag fixes deferred to P4 are NOT pulled forward (P3 stays scoped)
+- Owner demo script runs clean on the shipped code
+- No placeholder/stub text in owner-visible surfaces
+- The 3-click criterion still pinned green post-fixes
 
 ## Verification
 
-- Gates green; happy path walk passes; REQ-60, REQ-61 covered
-- 3-click criterion pinned by test
+- Gates: typecheck ✓ lint ✓ tests 42/42 ✓ build ✓ (re-run after P0 fixes, plus by audit)
+- border-hairline utility verified present in compiled production CSS
+- Review P0 worklist: 5/5 closed with tests where feasible (dials unit test, empty-question guard test)
 
 ## Scope Guard
 
-P3 only. The P4 final phase handles: review, audit, P1 flags (P1-1 drawer hyphen rendering from P1, P1-2 empty-question guard from P1, P2's routed P2 notes), owner approval gate, milestone merge. No v0.2 work.
\ No newline at end of file
+Milestone merge to main, v0.1.4 release publication, v0.2 unblocking, and all v0.2 P1 work wait on the owner approval gate (REQ-62, prototype_approval escalation hook). The pipeline does not proceed autonomously past this commit.
\ No newline at end of file
diff --git a/.ciagent/ROADMAP.md b/.ciagent/ROADMAP.md
index 45b08b5..eb6e284 100644
--- a/.ciagent/ROADMAP.md
+++ b/.ciagent/ROADMAP.md
@@ -38,7 +38,7 @@ Two milestones, owner-directed 2026-09-27. **Milestone v0.1 — Prototype**: the
 - **Requirements covered**: milestone gate (REQ-52..61 approved)
 - **Success criteria**: owner approval recorded; v0.1.4 tagged; milestone branches deleted per ship workflow
 - **Depends on**: 3
-- **Status**: not_started
+- **Status**: in_progress (review PASS-WITH-FIXES → 5 P0s fixed; audit PASS; awaiting owner approval)
 
 ## Milestone v0.2 — MVP (feature) — BLOCKED until prototype approval
 
diff --git a/README.md b/README.md
index f3b2fef..da9be12 100644
--- a/README.md
+++ b/README.md
@@ -22,7 +22,9 @@ Three clicks from cold start to initiated research.
 |---|---|
 | `/ask` | One question box + one optional context box. Nothing else competing for attention. |
 | `/choose` | All 8 locked methodology cards — mock ETA, cost estimate, why-recommended, and the 12 methodology dials behind each card's Advanced drawer. |
-| `/run/[id]` | Placeholder — Plan & Run ships in phase 2. |
+| `/run/[id]` | Plan & Run: approve the research plan (sub-questions, source universe, outline), watch the live trace stream, pause/resume or cancel. |
+| `/run/[id]/report` | Reading Room: the report with page-anchored citations — hover or focus any `[n]` marker for the exact quoted passage — plus the citation ledger and Markdown/JSON exports. |
+| `/library` | v2 preview: project graph and task inbox mockups rendered from fixture data. |
 
 All data flows through a typed `ApiClient` interface backed by `MockApiClient` — the same components will talk to the FastAPI backend via the OpenAPI-generated client in the MVP milestone, with zero rewrites.
 
diff --git a/frontend/README.md b/frontend/README.md
index 3ec3f6c..499420b 100644
--- a/frontend/README.md
+++ b/frontend/README.md
@@ -61,7 +61,7 @@ The happy path:
 
 ```
 src/
-  app/            App Router routes (ask, choose, run/[id], run/[id]/report)
+  app/            App Router routes (ask, choose, run/[id], run/[id]/report, library)
   components/     UI components (server by default, client where interactive)
   lib/            Domain types, ApiClient boundary, mock data, stores
   test/           Test harness setup
@@ -71,5 +71,5 @@ Prototype scope note: ETAs, cost estimates, and the scripted run are
 realistic fixture data — not live research. The trace stream replays a
 deterministic Deep Report run with pause/resume/cancel controls; the
 Reading Room's report, ledger, and exports are generated from the same
-fixture. The v2 surface mockups (project graph, task inbox) ship in the
-next phase.
\ No newline at end of file
+fixture. The `/library` project graph and task inbox are v2 preview
+mockups behind the same ApiClient boundary.
\ No newline at end of file
diff --git a/frontend/src/app/globals.css b/frontend/src/app/globals.css
index 4d97c74..065d95b 100644
--- a/frontend/src/app/globals.css
+++ b/frontend/src/app/globals.css
@@ -50,6 +50,7 @@
   --color-ink-soft: #4a4945;        /* secondary text              */
   --color-ink-faint: #8a8880;       /* muted metadata              */
   --color-line: #e5e2da;             /* hairline borders            */
+  --color-hairline: #e5e2da;        /* border-hairline utility      */
 
   --color-accent: #1e5c3f;           /* deep green — primary action */
   --color-accent-strong: #164732;    /* hover / pressed             */
diff --git a/frontend/src/components/choose-view.test.tsx b/frontend/src/components/choose-view.test.tsx
index 0fe962f..069e6b6 100644
--- a/frontend/src/components/choose-view.test.tsx
+++ b/frontend/src/components/choose-view.test.tsx
@@ -116,4 +116,26 @@ describe("ChooseView", () => {
       expect(runRoute).toMatch(/^\/run\/run-[a-z0-9]+$/);
     });
   });
+
+  it("empty-question guard: Start research disabled without a question (P1-2 fix)", async () => {
+    const methodologies = await loadFixtures();
+    const deepReport = methodologies.find((m) => m.slug === "deep-report");
+    if (!deepReport) throw new Error("fixture missing deep-report");
+
+    // Direct /choose navigation or a hard refresh: no carried question.
+    setAskState({ question: "" });
+    render(
+      <MethodologyCard methodology={deepReport} question={""} />,
+    );
+
+    const start = screen.getByRole("button", { name: "Start research" });
+    expect(start).toBeDisabled();
+    expect(start).toHaveAttribute("aria-disabled", "true");
+
+    // Whitespace-only questions are equally blocked.
+    render(
+      <MethodologyCard methodology={deepReport} question={"   "} />,
+    );
+    expect(screen.getAllByRole("button", { name: "Start research" })[0]).toBeDisabled();
+  });
 });
\ No newline at end of file
diff --git a/frontend/src/components/methodology-card.tsx b/frontend/src/components/methodology-card.tsx
index e893770..51d9e23 100644
--- a/frontend/src/components/methodology-card.tsx
+++ b/frontend/src/components/methodology-card.tsx
@@ -128,7 +128,13 @@ export function MethodologyCard({
         <button
           type="button"
           onClick={handleStart}
-          disabled={starting}
+          disabled={starting || question.trim() === ""}
+          aria-disabled={question.trim() === ""}
+          title={
+            question.trim() === ""
+              ? "Ask a question first — the Ask screen is one click away"
+              : undefined
+          }
           className="w-full rounded-pill bg-accent px-5 py-2.5 text-small font-medium text-paper-raised transition-colors hover:bg-accent-strong disabled:cursor-not-allowed disabled:opacity-50"
         >
           {starting ? "Starting…" : "Start research"}
diff --git a/frontend/src/components/reading-room.test.tsx b/frontend/src/components/reading-room.test.tsx
index aa942b4..00f0549 100644
--- a/frontend/src/components/reading-room.test.tsx
+++ b/frontend/src/components/reading-room.test.tsx
@@ -85,7 +85,7 @@ describe("ReadingRoom", () => {
     expect(tooltip.textContent).toContain(
       "Sulfide electrolytes exhibit ionic conductivity comparable to liquid electrolytes",
     );
-    expect(tooltip.textContent).toContain("arxiv.org");
+    expect(tooltip.textContent).toContain("example.org");
     expect(tooltip.textContent).toContain("tier 1");
   });
 
@@ -95,7 +95,7 @@ describe("ReadingRoom", () => {
     expect(markdown.filename).toBe(`researchx-${id}.md`);
     expect(markdown.content).toContain("## Sources");
     expect(markdown.content).toMatch(/\[.*\]\(https:\/\/.+\)/);
-    expect(markdown.content).toContain("](https://arxiv.org/abs/2401.00001)");
+    expect(markdown.content).toContain("](https://example.org/arxiv-sulfide-electrolytes-review)");
 
     const json = await api.exportReport(id, "json");
     expect(json.filename).toBe(`researchx-${id}.json`);
diff --git a/frontend/src/lib/api/mock/mock-client.test.ts b/frontend/src/lib/api/mock/mock-client.test.ts
index 63406f9..a289262 100644
--- a/frontend/src/lib/api/mock/mock-client.test.ts
+++ b/frontend/src/lib/api/mock/mock-client.test.ts
@@ -128,7 +128,7 @@ describe("MockApiClient (bound as `api`)", () => {
     // Per-claim markers survive as footnote-style references…
     expect(markdown.content).toContain("[^p1]:");
     // …and each citation is a real Markdown link to its source URL.
-    expect(markdown.content).toContain("](https://arxiv.org/abs/2401.00001)");
+    expect(markdown.content).toContain("](https://example.org/arxiv-sulfide-electrolytes-review)");
     expect(markdown.content).toContain("✓ live");
     expect(markdown.content).toContain("△ stale");
   });
diff --git a/frontend/src/lib/api/mock/scripted-run.ts b/frontend/src/lib/api/mock/scripted-run.ts
index 8563e87..f4bc918 100644
--- a/frontend/src/lib/api/mock/scripted-run.ts
+++ b/frontend/src/lib/api/mock/scripted-run.ts
@@ -36,9 +36,9 @@ export const scriptedCitations: readonly Citation[] = [
   },
   {
     id: "c2",
-    url: "https://arxiv.org/abs/2401.00001",
+    url: "https://example.org/arxiv-sulfide-electrolytes-review",
     title: "Sulfide Electrolytes for All-Solid-State Lithium Batteries: A Review",
-    site: "arxiv.org",
+    site: "example.org",
     passageId: "p2",
     quote:
       "Sulfide electrolytes exhibit ionic conductivity comparable to liquid electrolytes, but interface degradation during cycling remains the central obstacle to commercial cycle life.",
@@ -161,7 +161,7 @@ export function scriptedTraceEvents(runId: string): TraceEvent[] {
       type: "source_visited",
       message:
         "Visited: Sulfide Electrolytes for All-Solid-State Lithium Batteries: A Review",
-      detail: "arxiv.org · authority tier 1",
+      detail: "example.org · authority tier 1",
     },
     {
       id: `${runId}-e09`,
diff --git a/frontend/src/lib/dials.test.ts b/frontend/src/lib/dials.test.ts
new file mode 100644
index 0000000..31a0cb7
--- /dev/null
+++ b/frontend/src/lib/dials.test.ts
@@ -0,0 +1,35 @@
+/**
+ * Dial formatting tests (final-phase review fix — P0-1).
+ *
+ * formatDialValue must preserve separator meaning: hyphens stay
+ * hyphens ("per-claim" → "Per-Claim"), plus stays plus
+ * ("web+academic" → "Web+Academic"). This survived three phases
+ * unpinned; it is pinned now.
+ */
+
+import { describe, expect, it } from "vitest";
+import { formatDialValue } from "@/lib/dials";
+
+describe("formatDialValue", () => {
+  it("numbers pass through as strings", () => {
+    expect(formatDialValue(1)).toBe("1");
+    expect(formatDialValue(3)).toBe("3");
+  });
+
+  it("preserves hyphens: per-claim renders Per-Claim, not Per+Claim", () => {
+    expect(formatDialValue("per-claim")).toBe("Per-Claim");
+    expect(formatDialValue("plan-only")).toBe("Plan-Only");
+    expect(formatDialValue("plan+draft")).toBe("Plan+Draft");
+  });
+
+  it("preserves plus separators: web+academic renders Web+Academic", () => {
+    expect(formatDialValue("web+academic")).toBe("Web+Academic");
+  });
+
+  it("single words title-case; connectors stay lowercase", () => {
+    expect(formatDialValue("encyclopedic")).toBe("Encyclopedic");
+    // Space-separated words beyond connectors title-case normally;
+    // the connector rule keeps "and"/"or" lowercase mid-compound.
+    expect(formatDialValue("breadth and depth")).toBe("Breadth and depth");
+  });
+});
\ No newline at end of file
diff --git a/frontend/src/lib/dials.ts b/frontend/src/lib/dials.ts
index 7a1f97d..1add406 100644
--- a/frontend/src/lib/dials.ts
+++ b/frontend/src/lib/dials.ts
@@ -9,14 +9,21 @@ import { DIALS } from "@/lib/types";
 /** Humanize a raw dial value: numbers pass through, strings title-case. */
 export function formatDialValue(value: DialValues[keyof DialValues]): string {
   if (typeof value === "number") return String(value);
+  // Split into words AND separators, title-case words, keep the
+  // original separator characters ("per-claim" → "Per-Claim",
+  // "web+academic" → "Web+Academic") — separators are meaningful.
   return value
-    .split(/[-+]/)
+    .split(/([-+])/)
     .map((part) =>
-      part === "and" || part === "or"
+      part === "-" || part === "+"
         ? part
-        : part.charAt(0).toUpperCase() + part.slice(1),
+        : part === "and" || part === "or"
+          ? part
+          : part === ""
+            ? part
+            : part.charAt(0).toUpperCase() + part.slice(1),
     )
-    .join("+");
+    .join("");
 }
 
 /** All 12 dial definitions (for the Advanced drawer). */