feat: 接通实现核验修复到正向运行采纳

任务书16第二块：角色分派与每调用轻量检查，提交及代码原子记账，按次数核验修复且耗尽不跑，串行正式运行及恢复引用指标，采纳含null统一公开。451项免费测试通过，全新上下文独立审阅未发现问题；真实研究验收仍待进行。

tests/flow-docker.test.ts

@@ -1,9 +1,16 @@
 import assert from "node:assert/strict";
 import { spawnSync } from "node:child_process";
-import { cpSync, mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { cpSync, mkdirSync, mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
 import { tmpdir } from "node:os";
 import { join } from "node:path";
 import { test, type TestContext } from "node:test";
+import { lightChecker } from "../src/flow/task-code.ts";
+import { recover } from "../src/flow/recover.ts";
+import { containerName, freshDir, runPaths } from "../src/flow/paths.ts";
+import type { Study } from "../src/flow/study.ts";
+import { Board, protocolMetrics, type FrozenProtocol } from "../src/materials/index.ts";
+import { loadTask, formalRun as runDefinition, prepareRun } from "../src/tasks/index.ts";
+import { runBalance } from "../src/rules/index.ts";
 import { runStudy } from "../src/flow/index.ts";
 import { Ledger } from "../src/ledger/index.ts";
 import { Runner } from "../src/runner/index.ts";
@@ -76,3 +83,67 @@ test("docker run 没启动：写结果\"没启动\"后停下，原因带 Docker
   assert.match((ending as { reason: string }).reason, /运行 base-1 没有启动.*unknown or invalid runtime name: inforsi-test-no-such-runtime/);
   assert.deepEqual(ledgerKinds(dir), ["config", "code", "runPlanned", "runResult", "event"]);
 });
+
+
+/** 不派模型，只接真实任务镜像、账本和运行器，检查 flow 的两处 Docker 接线。 */
+async function runtimeStudy(t: TestContext, suffix: string): Promise<Study> {
+  const dir = tempDir(t);
+  const id = studyId(t, suffix);
+  const task = loadTask(TASK_DIR);
+  const runner = new Runner(null);
+  const image = await runner.buildImage(task.dir, `inforsi-${id}`);
+  const ledger = Ledger.open(join(dir, "ledger.sqlite"));
+  t.after(() => ledger.close());
+  return { dir, task, runner, ledger, stop: new AbortController().signal,
+    frozen: { config: { ...fullConfig(), studyId: id, lightChecks: { maxCount: 1, seconds: 7 } }, facts: task.facts, image, seed: 1 } } as Study;
+}
+
+test("flow 轻量检查：每次调用各自限次和限时，真实 Docker 只检查当前目录，不登记正式运行", { timeout: 120_000 }, async (t) => {
+  const s = await runtimeStudy(t, "light");
+  const work = join(s.dir, "work");
+  mkdirSync(work);
+  writeFileSync(join(work, "initial.py"), "raise RuntimeError('不应执行候选代码')\ndef run_packing():\n    pass\n");
+  for (const call of ["first-call", "second-call"]) {
+    freshDir(join(s.dir, "calls", call));
+    const check = lightChecker(s, call, work);
+    const result = JSON.parse(await check());
+    assert.deepEqual(result, { used: 1, maxCount: 1, seconds: 7, result: { passed: true } });
+    assert.match(await check(), /已用完（上限 1 次）/);
+    assert.equal(await s.runner.status(containerName(s.frozen.config.studyId, `${call}-check-1`)), "absent");
+  }
+  assert.deepEqual(s.ledger.records(), []);
+});
+
+test("正向恢复：接上原容器不重跑，补启动只扣一次，引用指标沿用原登记缺失作废规则", { timeout: 120_000 }, async (t) => {
+  const s = await runtimeStudy(t, "forward-recover");
+  const metric = { id: "old-c.required", meaning: "必测量", unit: "次", direction: "maximize" as const, howMeasured: "原登记定义", missing: "invalidateRun" as const };
+  const original: FrozenProtocol = { round: 0, baseline: "base", proposals: [], factors: [], safeguards: [], metrics: { defined: [metric], reused: [] }, clauses: [], body: "登记" };
+  const current: FrozenProtocol = { ...original, round: 1, metrics: { defined: [], reused: [metric.id] } };
+  s.ledger.append("protocol", { candidate: "old-c", protocol: original });
+  s.ledger.append("protocol", { candidate: "r1-c1", protocol: current });
+  const board = new Board(s.ledger.records());
+  const run = "forward-1";
+  const paths = runPaths(s.dir, run);
+  const definition = runDefinition(s.task, s.frozen.image, "1");
+  s.ledger.append("runPlanned", { run, purpose: "forward", round: 1, object: "r1-c1", definition });
+  freshDir(paths.base);
+  prepareRun(s.task, s.task.baselineDir, freshDir(paths.run), protocolMetrics(board, current));
+  const name = containerName(s.frozen.config.studyId, run);
+  t.after(() => { const result = docker(["rm", "-f", name]); if (result.status !== 0 && !result.stderr.includes("No such container")) throw new Error(result.stderr); });
+  assert.deepEqual(await s.runner.start(name, definition, paths.run), { started: true });
+  // 中断点：容器已启动，账本还没写 runStarted。恢复不调用 formalRun，而是找到原容器等它。
+  assert.equal(await recover(s), null);
+  const records = s.ledger.records();
+  assert.equal(records.filter((r) => r.kind === "runPlanned").length, 1);
+  assert.equal(records.filter((r) => r.kind === "runStarted").length, 1);
+  assert.equal(records.filter((r) => r.kind === "runResult").length, 1);
+  const result = records.find((r) => r.kind === "runResult")!;
+  if (result.kind !== "runResult") throw new Error("缺运行结果");
+  assert.equal(result.content.judged!.verdict, "invalid");
+  assert.match(result.content.judged!.verdictReason, /old-c.required/);
+  assert.equal(result.content.judged!.metrics[metric.id]!.status, "unmeasured");
+  assert.deepEqual(runBalance(13, new Board(records).runStates()), { charged: 1, reserved: 0, balance: 12 });
+  assert.equal(await s.runner.status(name), "absent");
+  assert.equal(await recover(s), null);
+  assert.equal(s.ledger.records().length, records.length);
+});