diff --git a/src/app/api/mental-health/assessments/__tests__/route.test.ts b/src/app/api/mental-health/assessments/__tests__/route.test.ts
new file mode 100644
index 000000000..40bdc9bea
--- /dev/null
+++ b/src/app/api/mental-health/assessments/__tests__/route.test.ts
@@ -0,0 +1,211 @@
+import { describe, it, expect, vi, beforeEach } from "vitest";
+import { NextRequest } from "next/server";
+
+/**
+ * v1.25.0 — `POST /api/mental-health/assessments` happy path.
+ *
+ * Regression guard for the dead-on-arrival bug: the handler used to pass the
+ * `safeJson` `{ data, error }` WRAPPER straight into `safeParse`, so the schema
+ * never matched and every POST returned 422. This pins that a valid body now
+ * (a) parses, (b) persists with a server-authoritative total / band / item-9
+ * flag, (c) writes the derived `*_SCORE` Measurement, and (d) surfaces the
+ * locale-aware crisis set on a positive item 9.
+ */
+
+vi.mock("@/lib/db", () => ({
+  prisma: {
+    appSettings: { findUnique: vi.fn().mockResolvedValue(null) },
+    mentalHealthAssessment: { create: vi.fn() },
+    measurement: { create: vi.fn().mockResolvedValue({}) },
+  },
+}));
+
+vi.mock("@/lib/modules/gate", async (importOriginal) => ({
+  ...(await importOriginal<typeof import("@/lib/modules/gate")>()),
+  requireModuleEnabled: vi.fn().mockResolvedValue({ enabled: true }),
+  resolveModuleMap: vi.fn().mockResolvedValue({}),
+}));
+
+vi.mock("@/lib/rate-limit", () => ({
+  checkRateLimit: vi
+    .fn()
+    .mockResolvedValue({ allowed: true, remaining: 29, resetAt: Date.now() }),
+  rateLimitHeaders: vi.fn().mockReturnValue({}),
+}));
+
+vi.mock("@/lib/ai/coach/bytes-codec", () => ({
+  encryptToBytes: vi.fn(() => Buffer.from("ciphertext")),
+}));
+
+vi.mock("@/lib/auth/session", () => ({ getSession: vi.fn() }));
+vi.mock("@/lib/auth/audit", () => ({
+  auditLog: vi.fn().mockResolvedValue(undefined),
+}));
+vi.mock("@/lib/logging/transports", () => ({ emitIfSampled: vi.fn() }));
+vi.mock("@/lib/db-compat", () => ({
+  ensureDbCompatibility: vi.fn().mockResolvedValue(undefined),
+}));
+vi.mock("next/headers", () => ({
+  headers: vi.fn(async () => ({ get: () => null })),
+  cookies: vi.fn(async () => ({
+    get: () => undefined,
+    set: () => {},
+    delete: () => {},
+  })),
+}));
+
+import { POST } from "../route";
+import { getSession } from "@/lib/auth/session";
+import { requireModuleEnabled } from "@/lib/modules/gate";
+import { checkRateLimit } from "@/lib/rate-limit";
+import { prisma } from "@/lib/db";
+
+const SESSION_OK = {
+  session: { id: "sess-1", expiresAt: new Date(Date.now() + 3_600_000) },
+  user: {
+    id: "user-1",
+    username: "testuser",
+    role: "USER" as const,
+    locale: "en",
+  },
+};
+
+const callPost = POST as unknown as (req: NextRequest) => Promise<Response>;
+
+function makeReq(body: unknown): NextRequest {
+  return new NextRequest(
+    new URL("http://localhost/api/mental-health/assessments"),
+    {
+      method: "POST",
+      headers: { "content-type": "application/json" },
+      body: JSON.stringify(body),
+    },
+  );
+}
+
+beforeEach(() => {
+  vi.clearAllMocks();
+  vi.mocked(getSession).mockResolvedValue(SESSION_OK as never);
+  vi.mocked(requireModuleEnabled).mockResolvedValue({ enabled: true });
+  vi.mocked(checkRateLimit).mockResolvedValue({
+    allowed: true,
+    remaining: 29,
+    resetAt: Date.now(),
+  });
+});
+
+describe("POST /api/mental-health/assessments", () => {
+  it("parses a valid PHQ-9 body and returns 201 with the server-computed total/band/flag", async () => {
+    // Item 9 = 2 → safety-flagged; total = 8*1 + 2 = 10 → moderate band.
+    vi.mocked(prisma.mentalHealthAssessment.create).mockResolvedValue({
+      id: "mha_1",
+      instrument: "PHQ9",
+      locale: "en",
+      version: "standard",
+      totalScore: 10,
+      severityBand: "moderate",
+      item9Flagged: true,
+      crisisShownAt: new Date("2026-06-28T00:00:00.000Z"),
+      takenAt: new Date("2026-06-28T00:00:00.000Z"),
+      createdAt: new Date("2026-06-28T00:00:00.000Z"),
+    } as never);
+
+    const res = await callPost(
+      makeReq({
+        instrument: "PHQ9",
+        items: [1, 1, 1, 1, 1, 1, 1, 1, 2],
+        locale: "en",
+      }),
+    );
+
+    expect(res.status).toBe(201);
+    const body = (await res.json()) as {
+      data: {
+        assessment: {
+          totalScore: number;
+          severityBand: string;
+          item9Flagged: boolean;
+        };
+        actionThreshold: number;
+        crisis: { emergencyNumber: string } | null;
+      };
+    };
+    expect(body.data.assessment.totalScore).toBe(10);
+    expect(body.data.assessment.severityBand).toBe("moderate");
+    expect(body.data.assessment.item9Flagged).toBe(true);
+    expect(body.data.actionThreshold).toBe(10);
+    // Positive item 9 → crisis resources surfaced immediately.
+    expect(body.data.crisis).not.toBeNull();
+    expect(body.data.crisis?.emergencyNumber).toBeTruthy();
+  });
+
+  it("persists server-authoritative scoring (userId from session, total recomputed) + the derived measurement", async () => {
+    vi.mocked(prisma.mentalHealthAssessment.create).mockResolvedValue({
+      id: "mha_2",
+      instrument: "GAD7",
+      locale: "en",
+      version: "standard",
+      totalScore: 6,
+      severityBand: "mild",
+      item9Flagged: false,
+      crisisShownAt: null,
+      takenAt: new Date("2026-06-28T00:00:00.000Z"),
+      createdAt: new Date("2026-06-28T00:00:00.000Z"),
+    } as never);
+
+    // GAD-7, total 1+2+0+1+1+1+0 = 6 → mild, no safety item.
+    const res = await callPost(
+      makeReq({
+        instrument: "GAD7",
+        items: [1, 2, 0, 1, 1, 1, 0],
+      }),
+    );
+    expect(res.status).toBe(201);
+
+    const createArg = vi.mocked(prisma.mentalHealthAssessment.create).mock
+      .calls[0][0] as {
+      data: {
+        userId: string;
+        instrument: string;
+        totalScore: number;
+        severityBand: string;
+        item9Flagged: boolean;
+      };
+    };
+    expect(createArg.data.userId).toBe("user-1");
+    expect(createArg.data.instrument).toBe("GAD7");
+    expect(createArg.data.totalScore).toBe(6);
+    expect(createArg.data.severityBand).toBe("mild");
+    expect(createArg.data.item9Flagged).toBe(false);
+
+    // Derived projection: one PHQ9_SCORE/GAD7_SCORE Measurement per assessment.
+    const measArg = vi.mocked(prisma.measurement.create).mock.calls[0][0] as {
+      data: { type: string; value: number; unit: string; externalId: string };
+    };
+    expect(measArg.data.type).toBe("GAD7_SCORE");
+    expect(measArg.data.value).toBe(6);
+    expect(measArg.data.unit).toBe("score");
+    expect(measArg.data.externalId).toBe("assessment:mha_2");
+  });
+
+  it("rejects a wrong item count with 422 and never writes", async () => {
+    const res = await callPost(
+      makeReq({ instrument: "PHQ9", items: [1, 2, 3] }), // 3 ≠ 9
+    );
+    expect(res.status).toBe(422);
+    expect(prisma.mentalHealthAssessment.create).not.toHaveBeenCalled();
+  });
+
+  it("returns the module 403 when the mental-health module is off", async () => {
+    const { apiError } = await import("@/lib/api-response");
+    vi.mocked(requireModuleEnabled).mockResolvedValue({
+      enabled: false,
+      response: apiError("disabled", 403, { errorCode: "module.disabled" }),
+    });
+    const res = await callPost(
+      makeReq({ instrument: "PHQ9", items: [0, 0, 0, 0, 0, 0, 0, 0, 0] }),
+    );
+    expect(res.status).toBe(403);
+    expect(prisma.mentalHealthAssessment.create).not.toHaveBeenCalled();
+  });
+});
diff --git a/src/lib/mental-health/__tests__/instruments.test.ts b/src/lib/mental-health/__tests__/instruments.test.ts
new file mode 100644
index 000000000..a97e92853
--- /dev/null
+++ b/src/lib/mental-health/__tests__/instruments.test.ts
@@ -0,0 +1,94 @@
+import { describe, it, expect } from "vitest";
+
+import {
+  INSTRUMENTS,
+  scoreTotal,
+  severityBand,
+  isSafetyFlagged,
+} from "../instruments";
+
+/**
+ * v1.25.0 — pin the highest-consequence numbers in the release: the PHQ-9 /
+ * GAD-7 totals, the validated severity-band edges (5 / 10 / 15 / 20 and
+ * 5 / 10 / 15), the ≥10 action threshold, and the item-9 ANY-non-zero safety
+ * flag. A future off-by-one on any band edge or the safety predicate must fail
+ * here. Ground truth: Kroenke 2001 (PHQ-9) / Spitzer 2006 (GAD-7), mirrored in
+ * `.planning/research/v125/clinical-instruments.md`.
+ */
+describe("mental-health instruments", () => {
+  describe("scoreTotal", () => {
+    it("sums the item answers", () => {
+      expect(scoreTotal([0, 1, 2, 3, 0, 1, 2, 3, 1])).toBe(13);
+      expect(scoreTotal([0, 0, 0, 0, 0, 0, 0])).toBe(0);
+    });
+
+    it("returns the max when every item is 3", () => {
+      expect(scoreTotal(Array(9).fill(3))).toBe(27); // PHQ-9 max
+      expect(scoreTotal(Array(7).fill(3))).toBe(21); // GAD-7 max
+    });
+  });
+
+  describe("severityBand — PHQ-9 edges (0–4 / 5–9 / 10–14 / 15–19 / 20–27)", () => {
+    const cases: Array<[number, string]> = [
+      [0, "minimal"],
+      [4, "minimal"],
+      [5, "mild"],
+      [9, "mild"],
+      [10, "moderate"],
+      [14, "moderate"],
+      [15, "modSevere"],
+      [19, "modSevere"],
+      [20, "severe"],
+      [27, "severe"],
+    ];
+    it.each(cases)("PHQ9 total %i → %s", (total, band) => {
+      expect(severityBand("PHQ9", total)).toBe(band);
+    });
+  });
+
+  describe("severityBand — GAD-7 edges (0–4 / 5–9 / 10–14 / 15–21)", () => {
+    const cases: Array<[number, string]> = [
+      [0, "minimal"],
+      [4, "minimal"],
+      [5, "mild"],
+      [9, "mild"],
+      [10, "moderate"],
+      [14, "moderate"],
+      [15, "severe"],
+      [21, "severe"],
+    ];
+    it.each(cases)("GAD7 total %i → %s", (total, band) => {
+      expect(severityBand("GAD7", total)).toBe(band);
+    });
+  });
+
+  describe("action threshold is ≥10 for both instruments", () => {
+    it("PHQ-9", () => {
+      expect(INSTRUMENTS.PHQ9.actionThreshold).toBe(10);
+    });
+    it("GAD-7", () => {
+      expect(INSTRUMENTS.GAD7.actionThreshold).toBe(10);
+    });
+  });
+
+  describe("isSafetyFlagged — PHQ-9 item 9 (index 8), ANY non-zero", () => {
+    it("flags when item 9 = 1, 2, or 3 regardless of total", () => {
+      for (const answer of [1, 2, 3]) {
+        const items = [0, 0, 0, 0, 0, 0, 0, 0, answer];
+        expect(isSafetyFlagged("PHQ9", items)).toBe(true);
+      }
+    });
+
+    it("does not flag when item 9 = 0 even with a high total", () => {
+      // Every other item maxed (24), item 9 = 0 → severe band, no flag.
+      const items = [3, 3, 3, 3, 3, 3, 3, 3, 0];
+      expect(severityBand("PHQ9", scoreTotal(items))).toBe("severe");
+      expect(isSafetyFlagged("PHQ9", items)).toBe(false);
+    });
+
+    it("never flags GAD-7 (no safety item)", () => {
+      expect(isSafetyFlagged("GAD7", [3, 3, 3, 3, 3, 3, 3])).toBe(false);
+      expect(INSTRUMENTS.GAD7.safetyItemIndex).toBeNull();
+    });
+  });
+});