diff --git a/docs/proposals/_BLOCKED-BY-CONSTRAINT.md b/docs/proposals/_BLOCKED-BY-CONSTRAINT.md
index da17ba56..eeec0a79 100644
--- a/docs/proposals/_BLOCKED-BY-CONSTRAINT.md
+++ b/docs/proposals/_BLOCKED-BY-CONSTRAINT.md
@@ -215,4 +215,4 @@ Linus 最后一句：**"先让数据结构对，其它 6 步会一起变简单"
 - No production mocks / fallbacks / silent catches — 仅 dev `tests/eval/` 例外。
 - No runtime logging / telemetry beyond existing electron-log usage.
 - No dual-write migrations without backup path。
-- `tests/eval/` 必须在 `electron-builder.yml` 的 files 过滤中排除，保证不进生产 bundle。
+- `electron-builder.yml` 采用 whitelist `files:` 模式（`out/**` + `package.json` + 指定 assets），`tests/eval/` 本就不在白名单内，无需新增排除规则。
diff --git a/tests/eval/assertion-judge.ts b/tests/eval/assertion-judge.ts
index 3f90d4cc..9ce46021 100644
--- a/tests/eval/assertion-judge.ts
+++ b/tests/eval/assertion-judge.ts
@@ -112,7 +112,9 @@ function evaluateOne(
     case "message_contains_substring": {
       for (const turn of turns) {
         if (turn.role !== assertion.role) continue;
-        const content = (turn as AssistantTurn | { content: unknown[] }).content;
+        // UserTurn and AssistantTurn both expose .content; the `role` guard
+        // above restricts us to those variants, so no cast is needed.
+        const content = turn.content;
         if (!Array.isArray(content)) continue;
         for (const block of content) {
           if (
diff --git a/tests/eval/harness.test.ts b/tests/eval/harness.test.ts
index 7a9b13d0..fffa4e1c 100644
--- a/tests/eval/harness.test.ts
+++ b/tests/eval/harness.test.ts
@@ -1,27 +1,42 @@
 /**
  * B5 eval harness — driver spec.
  *
- * This spec discovers every `*.jsonl` fixture under `tests/eval/fixtures/`
- * and runs it through the assertion judge. If there are zero fixtures, the
- * spec file has zero tests — vitest is invoked with `--passWithNoTests`
- * from the `test:eval` script so this is a green no-op until the user
- * records real fixtures.
+ * Two layers:
  *
- * The harness does NOT synthesize placeholder fixtures under any
- * circumstance. An empty `fixtures/` directory is a real state — the tier
- * is simply not yet exercising anything — and the green run reflects that
- * honestly.
+ * 1. Self-tests (always run). An in-memory fixture exercises the loader
+ *    contract, assertion judge, eval runner, and token ledger aggregation.
+ *    This gives `pnpm test:eval` real signal even before any `.jsonl`
+ *    fixture lands on disk. Zero fixture files = tier still exercised.
+ *
+ * 2. Fixture discovery (runs when `.jsonl` files exist). Every `*.jsonl`
+ *    under `tests/eval/fixtures/` is discovered and runs through the
+ *    judge. If the fixtures directory is missing we treat that as "no
+ *    fixtures yet", return an empty list, and let the self-tests carry
+ *    the signal on their own — we do NOT synthesize placeholder fixtures.
+ *
+ * Karpathy rule: bad fixture data must crash loud at the loader. The
+ * harness itself is validated by self-tests so a breakage there never
+ * hides behind an empty fixtures directory.
  */
 
+import { existsSync } from "node:fs";
 import { readdir } from "node:fs/promises";
 import { join } from "node:path";
 import { describe, expect, it } from "vitest";
-import { loadFixture } from "./fixture-loader.js";
+import { judgeAssertions } from "./assertion-judge.js";
 import { runEval } from "./eval-runner.js";
+import { loadFixture } from "./fixture-loader.js";
+import type {
+  Assertion,
+  AssistantTurn,
+  FixtureFile,
+  UserTurn,
+} from "./types.js";
 
 const FIXTURE_DIR = join(__dirname, "fixtures");
 
 async function listFixtures(): Promise<string[]> {
+  if (!existsSync(FIXTURE_DIR)) return [];
   const entries = await readdir(FIXTURE_DIR, { withFileTypes: true });
   return entries
     .filter((entry) => entry.isFile() && entry.name.endsWith(".jsonl"))
@@ -29,6 +44,108 @@ async function listFixtures(): Promise<string[]> {
     .sort();
 }
 
+function buildInMemoryFixture(assertions: Assertion[]): FixtureFile {
+  const userTurn: UserTurn = {
+    type: "turn",
+    turnIndex: 0,
+    role: "user",
+    content: [{ type: "text", text: "hire a writer" }],
+  };
+  const assistantTurn: AssistantTurn = {
+    type: "turn",
+    turnIndex: 1,
+    role: "assistant",
+    content: [
+      { type: "text", text: "Hiring a writer now." },
+      {
+        type: "toolCall",
+        id: "call_001",
+        name: "hire_agent",
+        arguments: { role: "writer", seniority: "senior" },
+      },
+    ],
+    tokenLedger: {
+      input: 120,
+      cache_read: 40,
+      cache_creation: 10,
+      output: 60,
+    },
+  };
+  return {
+    meta: {
+      type: "meta",
+      fixtureId: "self-test/in-memory",
+      piAgentCoreVersion: "0.0.0-self-test",
+      recordedAt: "2026-04-20T00:00:00.000Z",
+      scenario: "in-memory self-test fixture",
+      systemPromptHash: "sha256:self-test",
+    },
+    turns: [userTurn, assistantTurn],
+    assertions,
+  };
+}
+
+describe("harness self-test (in-memory fixture)", () => {
+  it("passes when every assertion matches the recorded trace", () => {
+    const fixture = buildInMemoryFixture([
+      { kind: "tool_called", name: "hire_agent", min_count: 1 },
+      {
+        kind: "tool_call_arg_matches",
+        name: "hire_agent",
+        arg_path: "role",
+        matches_regex: "^writer$",
+      },
+      { kind: "no_tool_called", name: "write_plan" },
+      {
+        kind: "message_contains_substring",
+        role: "assistant",
+        substring: "Hiring a writer",
+      },
+      { kind: "trace_max_turns", max: 5 },
+    ]);
+    const result = runEval(fixture);
+    expect(result.passed).toBe(true);
+    expect(result.fixtureId).toBe("self-test/in-memory");
+    expect(result.assertions).toHaveLength(5);
+    expect(result.assertions.every((r) => r.passed)).toBe(true);
+    expect(result.tokenLedger).toEqual({
+      input: 120,
+      cache_read: 40,
+      cache_creation: 10,
+      output: 60,
+    });
+    expect(result.durationMs).toBeGreaterThanOrEqual(0);
+  });
+
+  it("fails with a precise reason when an assertion misses", () => {
+    const fixture = buildInMemoryFixture([
+      { kind: "tool_called", name: "write_plan", min_count: 1 },
+      { kind: "no_tool_called", name: "hire_agent" },
+      { kind: "trace_max_turns", max: 1 },
+    ]);
+    const result = runEval(fixture);
+    expect(result.passed).toBe(false);
+    const reasons = result.assertions.map((r) => r.reason ?? "");
+    expect(reasons[0]).toContain('expected at least 1 call(s) to "write_plan"');
+    expect(reasons[1]).toContain('expected no call to "hire_agent"');
+    expect(reasons[2]).toContain("trace has 2 turns, max allowed is 1");
+  });
+
+  it("judgeAssertions is a pure function of (assertions, turns)", () => {
+    const fixture = buildInMemoryFixture([]);
+    const first = judgeAssertions(
+      [{ kind: "tool_called", name: "hire_agent" }],
+      fixture.turns,
+    );
+    const second = judgeAssertions(
+      [{ kind: "tool_called", name: "hire_agent" }],
+      fixture.turns,
+    );
+    expect(first).toEqual(second);
+    expect(first[0]?.passed).toBe(true);
+  });
+});
+
 const fixturePaths = await listFixtures();
 
 describe.each(fixturePaths)("eval fixture %s", (fixturePath) => {
diff --git a/tests/eval/replay-client.ts b/tests/eval/replay-client.ts
index aca37152..d642d549 100644
--- a/tests/eval/replay-client.ts
+++ b/tests/eval/replay-client.ts
@@ -90,13 +90,9 @@ export function createReplayStreamFn(fixture: FixtureFile): StreamFn {
         `replay-client: fixture "${fixture.meta.fixtureId}" exhausted after ${assistantTurns.length} assistant turns; agent requested turn ${cursor + 1}`,
       );
     }
-    const turn = assistantTurns[cursor];
+    // Cursor is bounds-checked above, so the element is guaranteed to exist.
+    const turn = assistantTurns[cursor]!;
     cursor += 1;
-    if (!turn) {
-      throw new Error(
-        `replay-client: internal cursor error at position ${cursor - 1}`,
-      );
-    }
     const message = buildAssistantMessage(
       turn,
       model.id,