feat: sync latest continuity hard-gate integration

README.md

@@ -15,34 +15,50 @@
 - 但沒有真的 dispatch 下一個 task
 - 最後流程卻還是收尾，造成 **auto-next break / continuity failure**
 
-目前 repo 內主要包含：
+## 目前已完成
 
-- continuity gate 規則與測試
+目前這個 repo 已經包含並驗證以下能力：
-- dispatch receipt binding 骨架與最小 receipt writer
-- anti-blackhole / delivery watchdog 的前置設計與基礎腳本
-- 相關 runbook / plan / state shape
 
-### 目前重點
+1. **continuity evaluator**
+   - task 完成、next action 已知、但沒有 next dispatch receipt，且 closure 狀態又不是 `waiting_user` / `blocked` / `pending_verification` 時，會判定 `continuity_failure`。
 
-這個 repo 目前著重於把以下能力拆成可測試、可逐步落地的切片：
+2. **dispatch receipt binding groundwork**
+   - 已有 continuity receipt storage 定義
+   - 已有最小 dispatch receipt writer
+   - 已有 continuity gate / dispatch binding 對應測試
 
-1. **continuity hard-gate**
+3. **`derivedAction` 與 `nextDerivedAction` 一致納入 continuity 判定**
-   - approved plan 內 task 完成後，若沒有 next dispatch receipt，且狀態又不是 `waiting_user` / `blocked` / `pending_verification`，則不應允許正常收尾。
+   - 不再只有 `nextDerivedAction` 才受 gate 約束。
 
-2. **dispatch receipt binding**
+4. **`dry_run_dispatch` 不得冒充真 receipt**
-   - 不只知道 planner 推導出下一步，而是要真的留下可驗證的 dispatch receipt。
+   - planner 的 dry-run 結果不再被 handler fallback 當成真實 dispatch receipt。
 
-3. **anti-blackhole groundwork**
+5. **fake receipt authority 已補強**
-   - 為後續的 subagent completion-delivery watchdog 提供 receipt / state / failure mode 基礎。
+   - continuity gate 不再接受任意 non-null `dispatchReceipt`
+   - 現在至少要求最小 receipt 欄位：
+     - `planId`
+     - `currentTask`
+     - `nextDerivedAction`
+     - `dispatchedAt`
 
-### 說明
+6. **hook integration 已接入**
+   - continuity gate 已接進 `hooks/force-recall/handler.ts`
+   - 目前會透過 `[APPROVED_PLAN_CONTINUITY_GATE]` block 注入現行 flow
 
-這是一個**聚焦匯出 repo**，不是整個原 workspace 的完整鏡像。目的是先把相關修補線獨立出來，方便：
+## 目前限制
 
-- 追蹤修補進度
+這條線雖然已經接入現行 flow，但目前仍偏向 **prompt-level hard-gate integration**，而不是 engine-level abort。也就是說：
-- 做 code review
+
-- 補 README / runbook / 測試
+- 已經不是只有規則文件
-- 後續再整合回主要流程
+- 已經不是只有獨立腳本測試
+- 但也還不是最底層 runtime/core 的絕對阻斷器
+
+## 下一步建議
+
+下一階段最合理的方向有兩條：
+
+1. **把 continuity hard-gate 再往更硬的 runtime enforcement 推進**
+2. **回頭補完 anti-blackhole / completion-delivery watchdog recovery 閉環**
 
 ---
 
@@ -61,31 +77,43 @@ The goal is to prevent this failure mode:
 - but the next task is never actually dispatched,
 - and the flow still closes out as if continuity were preserved.
 
-The repository currently includes:
+## Current State
 
-- continuity gate rules and tests
+The repo now includes and validates the following capabilities:
-- dispatch receipt binding skeletons and a minimal receipt writer
-- groundwork for anti-blackhole / delivery-watchdog handling
-- related runbooks, plans, and state-shape definitions
 
-### Current Focus
+1. **Continuity evaluator**
+   - When a task is complete, the next action is known, and there is no next dispatch receipt, and the closure state is not `waiting_user`, `blocked`, or `pending_verification`, the flow is classified as `continuity_failure`.
 
-This repo is currently structured to make the following capabilities testable and incrementally adoptable:
+2. **Dispatch receipt binding groundwork**
+   - continuity receipt storage shape
+   - minimal dispatch receipt writer
+   - continuity gate / dispatch binding tests
 
-1. **Continuity hard-gate**
+3. **`derivedAction` is treated as a real next-action source**
-   - If a task inside an approved plan is marked complete, and there is no next dispatch receipt, and the closure state is not `waiting_user`, `blocked`, or `pending_verification`, the flow should not be allowed to close normally.
+   - The gate no longer depends only on `nextDerivedAction`.
 
-2. **Dispatch receipt binding**
+4. **`dry_run_dispatch` is no longer accepted as a real receipt**
-   - It is not enough for the planner to derive a next action; that action must also produce verifiable dispatch evidence.
+   - Planner dry-run output is no longer promoted into a real dispatch receipt by handler fallback logic.
 
-3. **Anti-blackhole groundwork**
+5. **Fake receipt authority has been tightened**
-   - The repo also lays groundwork for future subagent completion-delivery watchdog logic through receipt/state/failure-mode definitions.
+   - The continuity gate no longer accepts any arbitrary non-null `dispatchReceipt`
+   - It now requires at least these minimum fields:
+     - `planId`
+     - `currentTask`
+     - `nextDerivedAction`
+     - `dispatchedAt`
 
-### Note
+6. **Hook integration is now present**
+   - The continuity gate is integrated into `hooks/force-recall/handler.ts`
+   - It currently enters the live flow through the `[APPROVED_PLAN_CONTINUITY_GATE]` injected block
 
-This is a **focused export repository**, not a full mirror of the original workspace. The intent is to isolate the relevant hardening work so it can be:
+## Current Limitation
 
-- reviewed more easily
+This workstream is now beyond pure documentation and beyond isolated script-level testing, but it still behaves more like a **prompt-level hard-gate integration** than a true engine-level abort mechanism.
-- iterated independently
+
-- documented clearly
+## Suggested Next Steps
-- integrated back into the main flow later
+
+Two reasonable follow-up directions remain:
+
+1. **push continuity hard-gate further toward stronger runtime enforcement**
+2. **return to anti-blackhole / completion-delivery watchdog recovery closure**

hooks/force-recall/handler.ts

@@ -8,6 +8,7 @@ const execFileAsync = promisify(execFile);
 const LONG_TASK_WRAPPER_TIMEOUT_MS = 8000;
 const LONG_TASK_GATE_LOCK_TIMEOUT_MS = 8000;
 const LONG_TASK_AUTO_CHAIN_PLANNER_TIMEOUT_MS = 8000;
+const APPROVED_PLAN_CONTINUITY_TIMEOUT_MS = 8000;
 
 type AutoChainPlanResult = {
   plannerStatus: string;
@@ -30,6 +31,14 @@ type GateLockResult = {
   allowedResponseModes?: string[];
 };
 
+type ApprovedPlanContinuityResult = {
+  ok: boolean;
+  status: string;
+  verdict: string;
+  reason?: string;
+  gate?: string;
+};
+
 function clamp(s: string, max = 1200): string {
   if (!s) return s;
   if (s.length <= max) return s;
@@ -328,6 +337,63 @@ async function runAutoChainPlanner(workspaceDir: string, gateLockResult: GateLoc
   return runJsonScript(plannerPath, workspaceDir, input, LONG_TASK_AUTO_CHAIN_PLANNER_TIMEOUT_MS);
 }
 
+function buildApprovedPlanContinuityInput(wrapperResult: any, autoChainPlanResult: AutoChainPlanResult | null): Record<string, unknown> | null {
+  if (!wrapperResult || wrapperResult.classification !== "long_task") return null;
+
+  const wrapperNextAction = wrapperResult?.nextDerivedAction ?? wrapperResult?.derivedAction ?? null;
+  const plannerDerivedAction = autoChainPlanResult?.derivedAction && autoChainPlanResult.derivedAction !== "none"
+    ? {
+        type: autoChainPlanResult.dispatchMode ?? "no_dispatch",
+        action: autoChainPlanResult.derivedAction,
+      }
+    : null;
+  const nextDerivedAction = wrapperNextAction ?? plannerDerivedAction;
+
+  if (nextDerivedAction == null) return null;
+
+  const replyClosureState = typeof wrapperResult?.replyClosureState === "string"
+    ? wrapperResult.replyClosureState
+    : (wrapperResult?.handoff?.mode === "button_path" ? "waiting_user" : "completed");
+
+  const dispatchReceipt = wrapperResult?.dispatchReceipt ?? null;
+
+  return {
+    planId: wrapperResult?.planId ?? "hook-preflight-approved-plan",
+    currentTask: wrapperResult?.currentTask ?? wrapperResult?.requiredNextAction ?? "hook-preflight-task",
+    taskState: wrapperResult?.taskState ?? (plannerDerivedAction ? "complete" : null),
+    nextDerivedAction,
+    replyClosureState,
+    dispatchReceipt,
+  };
+}
+
+async function runApprovedPlanContinuityGate(workspaceDir: string, wrapperResult: any, autoChainPlanResult: AutoChainPlanResult | null): Promise<ApprovedPlanContinuityResult | null> {
+  const continuityPath = path.join(workspaceDir, "scripts", "approved_plan_continuity_gate.mjs");
+  const input = buildApprovedPlanContinuityInput(wrapperResult, autoChainPlanResult);
+  if (!input) return null;
+  return runJsonScript(continuityPath, workspaceDir, input, APPROVED_PLAN_CONTINUITY_TIMEOUT_MS);
+}
+
+function buildApprovedPlanContinuityBlock(result: ApprovedPlanContinuityResult | null): string {
+  if (!result) return "";
+
+  const lines = [
+    "[APPROVED_PLAN_CONTINUITY_GATE]",
+    `status=${result.status}`,
+    `verdict=${result.verdict}`,
+  ];
+
+  if (result.reason) lines.push(`reason=${result.reason}`);
+
+  if (result.ok === false) {
+    lines.push("- HARD_GATE: Do not close out this reply as normal completion.");
+    lines.push("- HARD_GATE: Route back to continuity failure until a real next dispatch receipt exists, unless closure state is waiting_user, blocked, or pending_verification.");
+  }
+
+  lines.push("[/APPROVED_PLAN_CONTINUITY_GATE]", "");
+  return lines.join("\n");
+}
+
 function buildAutoChainPlanBlock(planResult: AutoChainPlanResult | null): string {
   if (!planResult) {
     return [
@@ -473,8 +539,11 @@ const forceRecall = async (event: any) => {
   ]);
   const gateLockResult = wrapperResult ? await runLongTaskGateLock(workspaceDir, wrapperResult) : null;
   const autoChainPlanResult = wrapperResult ? await runAutoChainPlanner(workspaceDir, gateLockResult, wrapperResult) : null;
+  const approvedPlanContinuityResult = wrapperResult
+    ? await runApprovedPlanContinuityGate(workspaceDir, wrapperResult, autoChainPlanResult)
+    : null;
 
-  if (!rulebook && !soul && !wrapperResult && !gateLockResult && !autoChainPlanResult) return;
+  if (!rulebook && !soul && !wrapperResult && !gateLockResult && !autoChainPlanResult && !approvedPlanContinuityResult) return;
 
   const wrapperBlock = wrapperResult
     ? [
@@ -500,6 +569,7 @@ const forceRecall = async (event: any) => {
 
   const gateLockBlock = buildGateLockBlock(gateLockResult);
   const autoChainPlanBlock = buildAutoChainPlanBlock(autoChainPlanResult);
+  const approvedPlanContinuityBlock = buildApprovedPlanContinuityBlock(approvedPlanContinuityResult);
 
   const recallBlock = [
     "[RECALL_GATE] Mandatory recall before ANY technical action/tool use.",
@@ -509,6 +579,7 @@ const forceRecall = async (event: any) => {
     wrapperBlock || null,
     gateLockBlock,
     autoChainPlanBlock,
+    approvedPlanContinuityBlock || null,
     rulebook ? `RULEBOOK (source: ${rulebookPath}):\n${clamp(rulebook, 1200)}` : null,
     soul ? `SOUL (source: ${soulPath}):\n${clamp(soul, 1200)}` : null,
     "[/RECALL_GATE]",

scripts/approved_plan_continuity_gate.mjs

@@ -3,6 +3,23 @@ import fs from 'node:fs';
 
 const LEGAL_TERMINAL_STATES = new Set(['waiting_user', 'blocked', 'pending_verification']);
 
+function isNonEmptyString(value) {
+  return typeof value === 'string' && value.trim().length > 0;
+}
+
+function isObject(value) {
+  return value != null && typeof value === 'object' && !Array.isArray(value);
+}
+
+function hasValidDispatchReceipt(receipt) {
+  if (!isObject(receipt)) return false;
+  if (!isNonEmptyString(receipt.planId)) return false;
+  if (!isNonEmptyString(receipt.currentTask)) return false;
+  if (!isObject(receipt.nextDerivedAction)) return false;
+  if (!isNonEmptyString(receipt.dispatchedAt)) return false;
+  return true;
+}
+
 function parseArgs(argv) {
   let inputPath = null;
   let compact = false;
@@ -59,7 +76,7 @@ function evaluateContinuity(payload) {
   const taskComplete = payload?.taskState === 'complete';
   const nextAction = payload?.nextDerivedAction ?? payload?.derivedAction ?? null;
   const nextActionKnown = nextAction != null;
-  const hasDispatchReceipt = payload?.dispatchReceipt != null;
+  const hasDispatchReceipt = hasValidDispatchReceipt(payload?.dispatchReceipt ?? null);
   const closureState = payload?.replyClosureState ?? null;
   const isLegalTerminalState = LEGAL_TERMINAL_STATES.has(closureState);
 

scripts/test_approved_plan_continuity_gate.mjs

@@ -149,13 +149,11 @@ const tests = [
         });
 
         if (result.status !== 0 && result.status !== null) {
-          throw new Error(`expected controlled execution, got status=${result.status}
+          throw new Error(`expected controlled execution, got status=${result.status}\n${result.stderr || result.stdout}`);
-${result.stderr || result.stdout}`);
         }
 
         if (!result.json || typeof result.json !== 'object') {
-          throw new Error(`expected JSON output
+          throw new Error(`expected JSON output\nstdout=${result.stdout}`);
-stdout=${result.stdout}`);
         }
 
         if (result.json.ok !== false) {
@@ -170,6 +168,54 @@ stdout=${result.stdout}`);
       }
     },
   },
+  {
+    name: 'continuity: fails when dispatchReceipt is a fake non-null object without minimum receipt fields',
+    run() {
+      const fixture = createFixture({
+        'input.json': {
+          planId: 'plan-fake-dispatch-receipt',
+          currentTask: 'task-6fake',
+          taskState: 'complete',
+          nextDerivedAction: {
+            type: 'message_subagent',
+            task: 'continue with task-7fake',
+          },
+          replyClosureState: 'completed',
+          dispatchReceipt: {
+            fake: true,
+          },
+        },
+      });
+
+      try {
+        const result = runGate({
+          args: ['--compact', '--input', fixture.path('input.json')],
+        });
+
+        if (result.status !== 0 && result.status !== null) {
+          throw new Error(`expected controlled execution, got status=${result.status}\n${result.stderr || result.stdout}`);
+        }
+
+        if (!result.json || typeof result.json !== 'object') {
+          throw new Error(`expected JSON output\nstdout=${result.stdout}`);
+        }
+
+        if (result.json.ok !== false) {
+          throw new Error(`expected continuity failure ok=false for fake dispatch receipt, got ${JSON.stringify(result.json)}`);
+        }
+
+        if (result.json.verdict !== 'continuity_failure') {
+          throw new Error(`expected verdict=continuity_failure for fake dispatch receipt, got ${JSON.stringify(result.json.verdict)}`);
+        }
+
+        if (result.json.reason !== 'missing_dispatch_receipt') {
+          throw new Error(`expected reason=missing_dispatch_receipt for fake dispatch receipt, got ${JSON.stringify(result.json.reason)}`);
+        }
+      } finally {
+        fixture.cleanup();
+      }
+    },
+  },
 
   {
     name: 'continuity: passes when task is complete, next action is known, and a dispatch receipt already exists',
@@ -202,13 +248,11 @@ stdout=${result.stdout}`);
         });
 
         if (result.status !== 0 && result.status !== null) {
-          throw new Error(`expected controlled execution, got status=${result.status}
+          throw new Error(`expected controlled execution, got status=${result.status}\n${result.stderr || result.stdout}`);
-${result.stderr || result.stdout}`);
         }
 
         if (!result.json || typeof result.json !== 'object') {
-          throw new Error(`expected JSON output
+          throw new Error(`expected JSON output\nstdout=${result.stdout}`);
-stdout=${result.stdout}`);
         }
 
         if (result.json.ok !== true) {
@@ -236,7 +280,7 @@ stdout=${result.stdout}`);
           dispatchReceipt: {
             planId: 'plan-derived-action-with-bound-dispatch',
             currentTask: 'task-6c',
-            derivedAction: {
+            nextDerivedAction: {
               type: 'message_subagent',
               task: 'continue with task-7c',
             },
@@ -251,13 +295,11 @@ stdout=${result.stdout}`);
         });
 
         if (result.status !== 0 && result.status !== null) {
-          throw new Error(`expected controlled execution, got status=${result.status}
+          throw new Error(`expected controlled execution, got status=${result.status}\n${result.stderr || result.stdout}`);
-${result.stderr || result.stdout}`);
         }
 
         if (!result.json || typeof result.json !== 'object') {
-          throw new Error(`expected JSON output
+          throw new Error(`expected JSON output\nstdout=${result.stdout}`);
-stdout=${result.stdout}`);
         }
 
         if (result.json.ok !== true) {
@@ -292,13 +334,11 @@ stdout=${result.stdout}`);
         });
 
         if (result.status !== 0 && result.status !== null) {
-          throw new Error(`expected controlled execution, got status=${result.status}
+          throw new Error(`expected controlled execution, got status=${result.status}\n${result.stderr || result.stdout}`);
-${result.stderr || result.stdout}`);
         }
 
         if (!result.json || typeof result.json !== 'object') {
-          throw new Error(`expected JSON output
+          throw new Error(`expected JSON output\nstdout=${result.stdout}`);
-stdout=${result.stdout}`);
         }
 
         if (result.json.ok !== true) {
@@ -333,13 +373,11 @@ stdout=${result.stdout}`);
         });
 
         if (result.status !== 0 && result.status !== null) {
-          throw new Error(`expected controlled execution, got status=${result.status}
+          throw new Error(`expected controlled execution, got status=${result.status}\n${result.stderr || result.stdout}`);
-${result.stderr || result.stdout}`);
         }
 
         if (!result.json || typeof result.json !== 'object') {
-          throw new Error(`expected JSON output
+          throw new Error(`expected JSON output\nstdout=${result.stdout}`);
-stdout=${result.stdout}`);
         }
 
         if (result.json.ok !== true) {
@@ -374,13 +412,11 @@ stdout=${result.stdout}`);
         });
 
         if (result.status !== 0 && result.status !== null) {
-          throw new Error(`expected controlled execution, got status=${result.status}
+          throw new Error(`expected controlled execution, got status=${result.status}\n${result.stderr || result.stdout}`);
-${result.stderr || result.stdout}`);
         }
 
         if (!result.json || typeof result.json !== 'object') {
-          throw new Error(`expected JSON output
+          throw new Error(`expected JSON output\nstdout=${result.stdout}`);
-stdout=${result.stdout}`);
         }
 
         if (result.json.ok !== true) {

scripts/test_force_recall_long_task_preflight.mjs

@@ -0,0 +1,489 @@
+#!/usr/bin/env node
+import assert from 'node:assert/strict';
+import fs from 'node:fs/promises';
+import os from 'node:os';
+import path from 'node:path';
+import { pathToFileURL } from 'node:url';
+import { execFile as execFileCallback } from 'node:child_process';
+import { promisify } from 'node:util';
+import { stripTypeScriptTypes } from 'node:module';
+
+const __dirname = path.dirname(new URL(import.meta.url).pathname);
+const repoRoot = path.resolve(__dirname, '..');
+const handlerPath = path.join(repoRoot, 'hooks', 'force-recall', 'handler.ts');
+const wrapperPath = path.join(repoRoot, 'scripts', 'long_task_governor_wrapper.mjs');
+const gateLockPath = path.join(repoRoot, 'scripts', 'long_task_gate_lock.mjs');
+const plannerPath = path.join(repoRoot, 'scripts', 'plan_long_task_auto_chain.mjs');
+const continuityGatePath = path.join(repoRoot, 'scripts', 'approved_plan_continuity_gate.mjs');
+const execFileAsync = promisify(execFileCallback);
+
+async function importTsModule(tsPath) {
+  const source = await fs.readFile(tsPath, 'utf8');
+  const jsSource = stripTypeScriptTypes(source, { mode: 'strip' });
+  const dataUrl = `data:text/javascript;charset=utf-8,${encodeURIComponent(jsSource)}\n//# sourceURL=${encodeURIComponent(pathToFileURL(tsPath).href)}`;
+  return import(dataUrl);
+}
+
+function escapeRegex(snippet) {
+  return snippet.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+}
+
+async function runScenario(forceRecall, requestText, workspaceDir = repoRoot) {
+  const event = {
+    type: 'message',
+    action: 'preprocessed',
+    context: {
+      workspaceDir,
+      body: requestText,
+      bodyForAgent: requestText,
+    },
+  };
+
+  await forceRecall(event);
+  const injected = event.context?.bodyForAgent;
+  assert.equal(typeof injected, 'string', 'event.context.bodyForAgent should be a string after handler runs');
+  return injected;
+}
+
+async function prepareTempWorkspace() {
+  const tempWorkspace = await fs.mkdtemp(path.join(os.tmpdir(), 'force-recall-workspace-'));
+  await fs.mkdir(path.join(tempWorkspace, 'scripts'), { recursive: true });
+  await fs.mkdir(path.join(tempWorkspace, 'hooks', 'force-recall'), { recursive: true });
+  await fs.mkdir(path.join(tempWorkspace, 'docs'), { recursive: true });
+
+  const copies = [
+    [wrapperPath, path.join(tempWorkspace, 'scripts', 'long_task_governor_wrapper.mjs')],
+    [gateLockPath, path.join(tempWorkspace, 'scripts', 'long_task_gate_lock.mjs')],
+    [plannerPath, path.join(tempWorkspace, 'scripts', 'plan_long_task_auto_chain.mjs')],
+    [continuityGatePath, path.join(tempWorkspace, 'scripts', 'approved_plan_continuity_gate.mjs')],
+    [handlerPath, path.join(tempWorkspace, 'hooks', 'force-recall', 'handler.ts')],
+    [path.join(repoRoot, 'docs', 'RULEBOOK.md'), path.join(tempWorkspace, 'docs', 'RULEBOOK.md')],
+    [path.join(repoRoot, 'SOUL.md'), path.join(tempWorkspace, 'SOUL.md')],
+  ];
+
+  for (const [src, dest] of copies) {
+    await fs.copyFile(src, dest);
+  }
+
+  return tempWorkspace;
+}
+
+async function withPatchedWrapper(tempContent, callback) {
+  const originalWrapper = await fs.readFile(wrapperPath, 'utf8');
+  await fs.writeFile(wrapperPath, tempContent, 'utf8');
+  try {
+    return await callback();
+  } finally {
+    await fs.writeFile(wrapperPath, originalWrapper, 'utf8');
+  }
+}
+
+async function withPatchedWrapperWorkspace(wrapperResult, callback) {
+  const tempWorkspace = await prepareTempWorkspace();
+  const wrapperScriptPath = path.join(tempWorkspace, 'scripts', 'long_task_governor_wrapper.mjs');
+  await fs.writeFile(wrapperScriptPath, buildWrapperScript(wrapperResult), 'utf8');
+
+  if (typeof wrapperResult.externalizedCheckpointPath === 'string' && wrapperResult.externalizedCheckpointPath.trim()) {
+    const checkpointPath = path.join(tempWorkspace, wrapperResult.externalizedCheckpointPath);
+    await fs.mkdir(path.dirname(checkpointPath), { recursive: true });
+    await fs.writeFile(checkpointPath, JSON.stringify({
+      kind: 'long_task_checkpoint',
+      currentStep: 'patched-wrapper-test',
+      nextStep: 'patched-wrapper-test-next',
+      verificationResult: 'checkpoint artifact readable in temp workspace',
+    }, null, 2) + '\n', 'utf8');
+  }
+
+  try {
+    return await callback(tempWorkspace);
+  } finally {
+    await fs.rm(tempWorkspace, { recursive: true, force: true });
+  }
+}
+
+function buildWrapperScript(wrapperResult) {
+  return `#!/usr/bin/env node\nprocess.stdout.write(JSON.stringify(${JSON.stringify(wrapperResult)}, null, 0) + "\\n");\n`;
+}
+
+async function main() {
+  await Promise.all([fs.access(wrapperPath), fs.access(gateLockPath), fs.access(plannerPath), fs.access(continuityGatePath)]);
+  const { default: forceRecall } = await importTsModule(handlerPath);
+  assert.equal(typeof forceRecall, 'function', 'force-recall handler should export default function');
+
+  const requestText = [
+    'Please inspect the workspace files and verify the hook injection path.',
+    'I need you to review the behavior, choose the final accept/reject decision,',
+    'and continue in background with a follow-up later.',
+  ].join(' ');
+  const plannerOnlyRequestText = [
+    'Please inspect the workspace files and verify the hook injection path.',
+    'Summarize the current dry-run planner state for technical inspection only.',
+  ].join(' ');
+
+  const checkpointWorkspace = await prepareTempWorkspace();
+  let realWrapperInjected;
+  try {
+    realWrapperInjected = await runScenario(forceRecall, 'Dispatch a subagent to inspect logs and wait for the result.', checkpointWorkspace);
+    const wrapperInputPath = path.join(checkpointWorkspace, 'wrapper-input.json');
+    await fs.writeFile(wrapperInputPath, JSON.stringify({
+      requestText: 'Dispatch a subagent to inspect logs and wait for the result.',
+      hasFilesOrSystems: false,
+      needsWaiting: false,
+      needsSubagent: false,
+      needsOwnerDecision: false,
+      canReplyNow: false,
+      taskName: 'Hook preflight classification',
+      currentStep: 'Classifying request at preprocessed hook',
+      nextStep: 'Carry governor recommendation into prompt context',
+      nextReportCondition: 'At next meaningful milestone',
+      waitingOn: 'none',
+      blocker: 'none',
+      checkpointTrigger: '',
+      externalizedTrigger: '',
+      triggerKind: '',
+    }), 'utf8');
+    const wrapperRaw = await fs.readFile(path.join(checkpointWorkspace, 'scripts', 'long_task_governor_wrapper.mjs'), 'utf8');
+    assert.ok(wrapperRaw.length > 0, 'temp workspace should contain wrapper script');
+    const { stdout: wrapperStdout } = await execFileAsync('node', [path.join(checkpointWorkspace, 'scripts', 'long_task_governor_wrapper.mjs'), '--compact', '--input', wrapperInputPath], { cwd: checkpointWorkspace, encoding: 'utf8' });
+    const wrapperOutput = JSON.parse(wrapperStdout);
+    const checkpointPath = path.join(checkpointWorkspace, wrapperOutput.externalizedCheckpointPath);
+    const checkpointBody = await fs.readFile(checkpointPath, 'utf8');
+    assert.ok(checkpointBody.trim().length > 0, 'real wrapper integration should emit readable checkpoint artifact');
+    assert.doesNotMatch(checkpointBody, /Hook preflight classification/, 'real wrapper artifact should not fall back to taskRecord.task_name');
+  } finally {
+    await fs.rm(checkpointWorkspace, { recursive: true, force: true });
+  }
+  assert.match(realWrapperInjected, /classification=long_task/, 'real wrapper integration should classify subagent wait as long_task');
+  assert.match(realWrapperInjected, /gateStatus=pass/, 'real wrapper integration should pass gate with real progress evidence');
+  assert.match(realWrapperInjected, /allowedResponseMode=silent_continuation/, 'real wrapper integration should preserve silent continuation allowance');
+  assert.doesNotMatch(realWrapperInjected, /reason=claimed progression without concrete progress evidence is forbidden/, 'real wrapper integration should not fail for missing progress evidence');
+  assert.doesNotMatch(realWrapperInjected, /requiredEvidence=progressEvidence/, 'real wrapper integration should not require synthetic progressEvidence repair');
+  assert.doesNotMatch(realWrapperInjected, /task_name/, 'real wrapper integration should not leak taskRecord.task_name fallback into gate/preflight text');
+
+  const injected = await runScenario(forceRecall, requestText);
+
+  const expectedSnippets = [
+    '[LONG_TASK_GOVERNOR_PREFLIGHT]',
+    'classification=long_task',
+    'silentLaunchOk=false',
+    'handoff.mode=button_path',
+    '[LONG_TASK_GATE_LOCK]',
+    'gateStatus=fail',
+    '[LONG_TASK_AUTO_CHAIN_PLAN]',
+    'plannerStatus=blocked_by_gate',
+    'derivedAction=none',
+    'dispatchMode=no_dispatch',
+    'autoChainAllowed=false',
+    'reason=gateStatus must pass before auto-chain planning can proceed',
+    'requiredEvidence=gateStatus=pass',
+    'requiredEvidence=externalizedCheckpoint',
+    'requiredEvidence=concreteNextAction',
+    'requiredEvidence=buttonPathMode',
+    'reason=silent long-task cannot continue without externalized checkpoint path',
+    'reason=claimed execution requires evidence of a concrete next action',
+    'reason=owner decision flow must end in button-path, not plain text',
+    'ENFORCEMENT: Hook input should include progressEvidence (or equivalent concrete fields) whenever a progression claim is present.',
+    'HARD_GATE: Block any plain-text handoff or silent-continuation claim when externalized checkpoint evidence is missing.',
+    'HARD_GATE: If owner decision is involved, do not replace button-path closure with plain-text handoff.',
+    'ENFORCEMENT: Forbidden path: plain-text handoff that pretends the long task is already continuing without an externalized checkpoint.',
+    'ENFORCEMENT: Forbidden path: stating you have already entered the next task/step when the record only contains planning language and no concrete execution evidence.',
+  ];
+
+  const unexpectedSnippets = [
+    'reason=claimed progression without concrete progress evidence is forbidden',
+    'requiredEvidence=progressEvidence',
+  ];
+
+  for (const snippet of expectedSnippets) {
+    assert.match(injected, new RegExp(escapeRegex(snippet)), `missing snippet: ${snippet}`);
+  }
+  for (const snippet of unexpectedSnippets) {
+    assert.doesNotMatch(injected, new RegExp(escapeRegex(snippet)), `unexpected snippet present: ${snippet}`);
+  }
+
+  const { evaluateGate } = await import(pathToFileURL(gateLockPath).href + `?t=${Date.now()}`);
+  assert.equal(typeof evaluateGate, 'function', 'long_task_gate_lock should export evaluateGate for direct tests');
+
+  const passResult = evaluateGate({
+    classification: 'long_task',
+    claimedExecution: true,
+    concreteNextAction: 'dispatch_follow_up_subagent',
+    autoChainNextAction: 'dispatch_follow_up_subagent',
+    autoChainDispatchEvidence: {
+      action: 'dispatch_follow_up_subagent',
+      dispatched: true,
+      event: 'dispatch',
+    },
+    progressionClaim: 'already progressing to the next step in background',
+    progressEvidence: { sessionKey: 'task-123' },
+  });
+  assert.equal(passResult.gateStatus, 'pass', 'pass-path should pass with concrete progressEvidence');
+
+  const failResult = evaluateGate({
+    classification: 'long_task',
+    claimedExecution: true,
+    concreteNextAction: 'dispatch_follow_up_subagent',
+    autoChainNextAction: 'dispatch_follow_up_subagent',
+    progressionClaim: 'already progressing to the next step in background',
+    executionEvidence: { concreteNextAction: 'dispatch_follow_up_subagent' },
+  });
+  assert.equal(failResult.gateStatus, 'fail', 'fail-path should fail when explicit auto-chain action lacks dispatch evidence');
+  assert.match(JSON.stringify(failResult), /autoChainDispatchEvidence/, 'fail-path should require autoChainDispatchEvidence');
+
+  const neutralResult = evaluateGate({
+    classification: 'long_task',
+    claimedExecution: true,
+    concreteNextAction: 'summarize findings for reply',
+    executionEvidence: { concreteNextAction: 'summarize findings for reply' },
+  });
+  assert.equal(neutralResult.gateStatus, 'pass', 'neutral-path should pass when there is no explicit auto-chain next action');
+  assert.doesNotMatch(JSON.stringify(neutralResult), /autoChainDispatchEvidence/, 'neutral-path should not require auto-chain dispatch evidence');
+
+  const directAutoChainFailResult = evaluateGate({
+    classification: 'long_task',
+    claimedExecution: true,
+    concreteNextAction: 'dispatch_follow_up_subagent',
+    autoChainNextAction: 'dispatch_follow_up_subagent',
+  });
+  assert.equal(directAutoChainFailResult.gateStatus, 'fail', 'direct evaluator should fail when explicit auto-chain action has no dispatch evidence');
+  assert.match(JSON.stringify(directAutoChainFailResult), /explicit auto-chain next action requires dispatched-action evidence/, 'direct evaluator fail-path should mention missing dispatched-action evidence');
+
+  const mismatchedDispatchEvidenceResult = evaluateGate({
+    classification: 'long_task',
+    claimedExecution: true,
+    concreteNextAction: 'dispatch_follow_up_subagent',
+    autoChainNextAction: 'dispatch_follow_up_subagent',
+    autoChainDispatchEvidence: {
+      action: 'dispatch_other_subagent',
+      dispatched: true,
+      event: 'dispatch',
+    },
+  });
+  assert.equal(mismatchedDispatchEvidenceResult.gateStatus, 'fail', 'mismatched dispatch evidence should fail');
+  assert.match(JSON.stringify(mismatchedDispatchEvidenceResult), /autoChainDispatchEvidence/, 'mismatched dispatch evidence should still require matching autoChainDispatchEvidence');
+
+  const fakeCheckpointDispatchEvidenceResult = evaluateGate({
+    classification: 'long_task',
+    claimedExecution: true,
+    concreteNextAction: 'dispatch_follow_up_subagent',
+    autoChainNextAction: 'dispatch_follow_up_subagent',
+    autoChainDispatchEvidence: {
+      sessionKey: 'task-123',
+      checkpointPath: 'checkpoints/task-123.json',
+    },
+  });
+  assert.equal(fakeCheckpointDispatchEvidenceResult.gateStatus, 'fail', 'checkpoint/session-only dispatch evidence should fail');
+  assert.match(JSON.stringify(fakeCheckpointDispatchEvidenceResult), /explicit auto-chain next action requires dispatched-action evidence/, 'checkpoint/session-only dispatch evidence should be rejected as fake dispatch evidence');
+
+  const neutralSnakeCaseResult = evaluateGate({
+    classification: 'long_task',
+    claimedExecution: true,
+    concreteNextAction: 'summarize findings for reply',
+    autoChainNextAction: 'checkpoint_session_metadata_only',
+    executionEvidence: { concreteNextAction: 'summarize findings for reply' },
+  });
+  assert.equal(neutralSnakeCaseResult.gateStatus, 'pass', 'neutral snake_case non-dispatch action should not trigger dispatch-evidence requirement');
+  assert.doesNotMatch(JSON.stringify(neutralSnakeCaseResult), /autoChainDispatchEvidence/, 'neutral snake_case non-dispatch action should not mention dispatch-evidence requirement');
+
+  const passInjected = await withPatchedWrapperWorkspace({
+    classification: 'long_task',
+    silentCandidate: true,
+    needsCheckpoint: true,
+    needsSubagent: false,
+    needsOwnerDecision: false,
+    silentLaunchOk: true,
+    silentLaunchReason: 'checkpoint established',
+    requiredNextAction: 'dispatch_follow_up_subagent',
+    autoChainDispatchEvidence: {
+      action: 'dispatch_follow_up_subagent',
+      dispatched: true,
+      event: 'dispatch',
+    },
+    progressEvidence: { sessionKey: 'task-123' },
+    externalizedCheckpointPath: 'checkpoints/task-123.json',
+    handoff: { mode: 'direct_reply' },
+  }, async (workspaceDir) => runScenario(forceRecall, requestText, workspaceDir));
+  assert.match(passInjected, /gateStatus=pass/, 'hook pass-path should pass when wrapper provides concrete progressEvidence');
+  assert.match(passInjected, /\[LONG_TASK_AUTO_CHAIN_PLAN\]/, 'hook pass-path should emit auto-chain plan block');
+  assert.match(passInjected, /plannerStatus=pass/, 'hook pass-path should expose planner pass result');
+  assert.match(passInjected, /derivedAction=dispatch_spec_review/, 'hook pass-path should derive dry-run spec review dispatch');
+  assert.match(passInjected, /dispatchMode=dry_run_dispatch/, 'hook pass-path should stay in dry-run dispatch mode');
+  assert.match(passInjected, /autoChainAllowed=true/, 'hook pass-path should allow auto-chain in dry-run planner output');
+  assert.match(passInjected, /\[APPROVED_PLAN_CONTINUITY_GATE\]/, 'hook pass-path should emit approved-plan continuity gate block');
+  assert.match(passInjected, /status=continuity_failure/, 'hook pass-path should fail continuity when planner only returns dry-run dispatch without a real receipt');
+  assert.match(passInjected, /verdict=continuity_failure/, 'hook pass-path should expose continuity failure verdict when no real dispatch receipt exists');
+  assert.match(passInjected, /reason=missing_dispatch_receipt/, 'hook pass-path should require a real dispatch receipt instead of treating dry-run dispatch as one');
+  assert.match(passInjected, /Route back to continuity failure until a real next dispatch receipt exists/, 'hook pass-path should hard-gate normal closeout until a real receipt exists');
+  assert.doesNotMatch(passInjected, /\[APPROVED_PLAN_CONTINUITY_GATE\][\s\S]*status=pass/, 'hook pass-path should not let approved-plan continuity pass on dry-run dispatch alone');
+
+  const failInjected = await withPatchedWrapper(buildWrapperScript({
+    classification: 'long_task',
+    silentCandidate: false,
+    needsCheckpoint: false,
+    needsSubagent: false,
+    needsOwnerDecision: false,
+    silentLaunchOk: false,
+    requiredNextAction: 'dispatch_follow_up_subagent',
+    handoff: { mode: 'direct_reply' },
+  }), async () => runScenario(forceRecall, requestText));
+  assert.match(failInjected, /gateStatus=fail/, 'hook fail-path should fail when wrapper exposes explicit auto-chain action without dispatch evidence');
+  assert.match(failInjected, /\[LONG_TASK_AUTO_CHAIN_PLAN\]/, 'hook fail-path should emit auto-chain plan block');
+  assert.match(failInjected, /plannerStatus=blocked_by_gate/, 'hook fail-path should report planner blocked by gate');
+  assert.match(failInjected, /derivedAction=none/, 'hook fail-path should not derive a dry-run action');
+  assert.match(failInjected, /dispatchMode=no_dispatch/, 'hook fail-path should remain no-dispatch');
+  assert.match(failInjected, /autoChainAllowed=false/, 'hook fail-path should not allow auto-chain');
+  assert.match(failInjected, /reason=explicit auto-chain next action requires dispatched-action evidence/, 'hook fail-path should mention missing dispatched-action evidence');
+  assert.match(failInjected, /requiredEvidence=autoChainDispatchEvidence/, 'hook fail-path should require autoChainDispatchEvidence');
+
+  const neutralInjected = await withPatchedWrapper(buildWrapperScript({
+    classification: 'long_task',
+    silentCandidate: false,
+    needsCheckpoint: false,
+    needsSubagent: false,
+    needsOwnerDecision: false,
+    silentLaunchOk: false,
+    requiredNextAction: 'summarize findings for reply',
+    handoff: { mode: 'direct_reply' },
+  }), async () => runScenario(forceRecall, requestText));
+  assert.match(neutralInjected, /gateStatus=pass/, 'hook neutral-path should pass when wrapper does not expose an explicit auto-chain action');
+  assert.match(neutralInjected, /\[LONG_TASK_AUTO_CHAIN_PLAN\]/, 'hook neutral-path should emit auto-chain plan block');
+  assert.match(neutralInjected, /plannerStatus=none/, 'hook neutral-path should report no derived auto-chain action');
+  assert.match(neutralInjected, /derivedAction=none/, 'hook neutral-path should keep derivedAction as none');
+  assert.match(neutralInjected, /dispatchMode=no_dispatch/, 'hook neutral-path should remain no-dispatch');
+  assert.match(neutralInjected, /autoChainAllowed=false/, 'hook neutral-path should keep auto-chain disabled');
+  assert.doesNotMatch(neutralInjected, /reason=explicit auto-chain next action requires dispatched-action evidence/, 'hook neutral-path should not fail on auto-chain evidence when no explicit tool action exists');
+
+  const fakeProgressEvidenceInjected = await withPatchedWrapper(buildWrapperScript({
+    classification: 'long_task',
+    silentCandidate: true,
+    needsCheckpoint: true,
+    needsSubagent: false,
+    needsOwnerDecision: false,
+    silentLaunchOk: true,
+    silentLaunchReason: 'task name exists but no externalized artifact',
+    taskRecord: { task_name: 'descriptive-task-name-only' },
+    handoff: { mode: 'direct_reply' },
+  }), async () => runScenario(forceRecall, requestText));
+  assert.match(fakeProgressEvidenceInjected, /gateStatus=fail/, 'hook fake-progress-evidence path should fail when only task_name exists');
+  assert.match(fakeProgressEvidenceInjected, /reason=claimed progression without concrete progress evidence is forbidden/, 'hook fake-progress-evidence path should mention missing concrete progress evidence');
+  assert.match(fakeProgressEvidenceInjected, /requiredEvidence=progressEvidence/, 'hook fake-progress-evidence path should require progressEvidence');
+  assert.match(fakeProgressEvidenceInjected, /reason=silent long-task cannot continue without externalized checkpoint path/, 'hook fake-progress-evidence path should also require real checkpoint evidence');
+
+  const specReviewWithoutEvidenceInjected = await withPatchedWrapper(buildWrapperScript({
+    classification: 'long_task',
+    silentCandidate: false,
+    needsCheckpoint: false,
+    needsSubagent: false,
+    needsOwnerDecision: false,
+    silentLaunchOk: true,
+    requiredNextAction: 'dispatch_code_quality_review',
+    autoChainDispatchEvidence: {
+      action: 'dispatch_code_quality_review',
+      dispatched: true,
+      event: 'dispatch',
+    },
+    progressEvidence: { sessionKey: 'task-spec-review-missing-evidence' },
+    externalizedCheckpointPath: 'checkpoints/task-spec-review-missing-evidence.json',
+    handoff: { mode: 'direct_reply' },
+  }), async () => runScenario(forceRecall, plannerOnlyRequestText));
+  assert.match(specReviewWithoutEvidenceInjected, /\[LONG_TASK_AUTO_CHAIN_PLAN\]/, 'hook spec-review missing-evidence path should emit auto-chain plan block');
+  assert.match(specReviewWithoutEvidenceInjected, /plannerStatus=blocked_by_evidence/, 'hook spec-review missing-evidence path should block on missing evidence');
+  assert.match(specReviewWithoutEvidenceInjected, /derivedAction=none/, 'hook spec-review missing-evidence path should not derive a dry-run action');
+  assert.match(specReviewWithoutEvidenceInjected, /dispatchMode=no_dispatch/, 'hook spec-review missing-evidence path should stay no-dispatch');
+  assert.match(specReviewWithoutEvidenceInjected, /autoChainAllowed=false/, 'hook spec-review missing-evidence path should not allow auto-chain');
+  assert.match(specReviewWithoutEvidenceInjected, /reason=review pass evidence missing for code quality review transition/, 'hook spec-review missing-evidence path should mention missing review evidence');
+  assert.match(specReviewWithoutEvidenceInjected, /requiredEvidence=reviewEvidence/, 'hook spec-review missing-evidence path should require reviewEvidence');
+
+  const fixSliceWithoutEvidenceInjected = await withPatchedWrapper(buildWrapperScript({
+    classification: 'long_task',
+    silentCandidate: false,
+    needsCheckpoint: false,
+    needsSubagent: false,
+    needsOwnerDecision: false,
+    silentLaunchOk: true,
+    silentLaunchReason: 'review blocked by findings',
+    requiredNextAction: 'dispatch_fix_slice',
+    autoChainDispatchEvidence: {
+      action: 'dispatch_fix_slice',
+      dispatched: true,
+      event: 'dispatch',
+    },
+    progressEvidence: { sessionKey: 'task-fix-slice-missing-evidence' },
+    externalizedCheckpointPath: 'checkpoints/task-fix-slice-missing-evidence.json',
+    handoff: { mode: 'direct_reply' },
+  }), async () => runScenario(forceRecall, plannerOnlyRequestText));
+  assert.match(fixSliceWithoutEvidenceInjected, /\[LONG_TASK_AUTO_CHAIN_PLAN\]/, 'hook fix-slice missing-evidence path should emit auto-chain plan block');
+  assert.match(fixSliceWithoutEvidenceInjected, /plannerStatus=blocked_by_evidence/, 'hook fix-slice missing-evidence path should block on missing evidence');
+  assert.match(fixSliceWithoutEvidenceInjected, /derivedAction=none/, 'hook fix-slice missing-evidence path should not derive a dry-run action');
+  assert.match(fixSliceWithoutEvidenceInjected, /dispatchMode=no_dispatch/, 'hook fix-slice missing-evidence path should stay no-dispatch');
+  assert.match(fixSliceWithoutEvidenceInjected, /autoChainAllowed=false/, 'hook fix-slice missing-evidence path should not allow auto-chain');
+  assert.match(fixSliceWithoutEvidenceInjected, /reason=blocker evidence missing for retry\/fix transition/, 'hook fix-slice missing-evidence path should mention missing blocker evidence');
+  assert.match(fixSliceWithoutEvidenceInjected, /requiredEvidence=blockerEvidence/, 'hook fix-slice missing-evidence path should require blockerEvidence');
+
+  const specReviewWithoutImplementationEvidenceInjected = await withPatchedWrapper(buildWrapperScript({
+    classification: 'long_task',
+    silentCandidate: false,
+    needsCheckpoint: false,
+    needsSubagent: false,
+    needsOwnerDecision: false,
+    silentLaunchOk: true,
+    requiredNextAction: 'dispatch_spec_review',
+    autoChainDispatchEvidence: {
+      action: 'dispatch_spec_review',
+      dispatched: true,
+      event: 'dispatch',
+    },
+    progressEvidence: { sessionKey: 'task-implementation-missing-evidence' },
+    externalizedCheckpointPath: 'checkpoints/task-implementation-missing-evidence.json',
+    handoff: { mode: 'direct_reply' },
+  }), async () => runScenario(forceRecall, plannerOnlyRequestText));
+  assert.match(specReviewWithoutImplementationEvidenceInjected, /\[LONG_TASK_AUTO_CHAIN_PLAN\]/, 'hook implementation missing-evidence path should emit auto-chain plan block');
+  assert.match(specReviewWithoutImplementationEvidenceInjected, /plannerStatus=blocked_by_evidence/, 'hook implementation missing-evidence path should block on missing evidence');
+  assert.match(specReviewWithoutImplementationEvidenceInjected, /derivedAction=none/, 'hook implementation missing-evidence path should not derive a dry-run action');
+  assert.match(specReviewWithoutImplementationEvidenceInjected, /dispatchMode=no_dispatch/, 'hook implementation missing-evidence path should stay no-dispatch');
+  assert.match(specReviewWithoutImplementationEvidenceInjected, /autoChainAllowed=false/, 'hook implementation missing-evidence path should not allow auto-chain');
+  assert.match(specReviewWithoutImplementationEvidenceInjected, /reason=implementation evidence missing for review-required next action/, 'hook implementation missing-evidence path should mention missing implementation evidence');
+  assert.match(specReviewWithoutImplementationEvidenceInjected, /requiredEvidence=executionEvidence/, 'hook implementation missing-evidence path should require executionEvidence');
+
+  const originalGateLock = await fs.readFile(gateLockPath, 'utf8');
+  const tempDir = await fs.mkdtemp(path.join(os.tmpdir(), 'force-recall-gate-lock-'));
+  const backupPath = path.join(tempDir, path.basename(gateLockPath));
+  await fs.writeFile(backupPath, originalGateLock, 'utf8');
+  await fs.writeFile(gateLockPath, '#!/usr/bin/env node\nprocess.exit(1);\n', 'utf8');
+
+  let degradedInjected;
+  try {
+    degradedInjected = await runScenario(forceRecall, requestText);
+  } finally {
+    const backup = await fs.readFile(backupPath, 'utf8');
+    await fs.writeFile(gateLockPath, backup, 'utf8');
+    await fs.rm(tempDir, { recursive: true, force: true });
+  }
+
+  const degradedExpectedSnippets = [
+    '[LONG_TASK_GATE_LOCK]',
+    'gateStatus=degraded',
+    'gateRequired=unknown',
+    'HARD_GATE: Evaluator unavailable is not permission to claim silent continuation or next-task progression without verifiable progress evidence.',
+    'HARD_GATE: Fall back to a non-silent, evidence-preserving follow-up if you cannot prove checkpoint state or concrete execution.',
+  ];
+
+  for (const snippet of degradedExpectedSnippets) {
+    assert.match(degradedInjected, new RegExp(escapeRegex(snippet)), `missing degraded snippet: ${snippet}`);
+  }
+
+  process.stdout.write(JSON.stringify({
+    ok: true,
+    gatePaths: {
+      pass: passResult.gateStatus,
+      fail: failResult.gateStatus,
+      neutral: neutralResult.gateStatus,
+    },
+    bodyPreview: injected.split('\n').slice(0, 35),
+  }, null, 2) + '\n');
+}
+
+main().catch((error) => {
+  console.error(error);
+  process.exitCode = 1;
+});