diff --git a/apps/api/src/modules/import/content-writer.ts b/apps/api/src/modules/import/content-writer.ts
index 13283aa..da66d6f 100644
--- a/apps/api/src/modules/import/content-writer.ts
+++ b/apps/api/src/modules/import/content-writer.ts
@@ -50,7 +50,7 @@ import { UNIVERSAL_ROW_VERSION, type UniversalRow } from '@sf/contracts'
 import type { AdapterEvent } from '@sf/adapter-wp'
 import { canonicalJson } from '../../common/canonical.js'
 import { contentEntries, contentRevisions, entryLocales, sourceRecords } from '../../db/schema.js'
-import type { ImportDb, ImportPipeline, ImportPipelineOutcome, ImportVersionBlock } from './import.service.js'
+import type { ImportDb, ImportPipeline, ImportPipelineOutcome, ImportVersionBlock, WriterContext } from './import.service.js'
 
 /** 事务句柄类型（db.transaction 回调参数——从 ImportDb 的事务签名提取，避免重复泛型展开） */
 type Tx = Parameters<Parameters<ImportDb['transaction']>[0]>[0]
@@ -149,18 +149,15 @@ interface ContentWriterCounts {
  * - contentEntries：新建 content_entries 数（幂等重跑/变化路径为 0）；
  * - revisionsCreated：新建 content_revisions 数（checksum 命中为 0——G2 验收锚点）；
  * - sourceRecords：新建 source_records 数（内容/attachment/unsupported/error 全计；
- *   跨快照重导按快照补行故可为正）；
+ *   跨快照重导按快照补行故可为正；NULL-entry 占位行收编**不计**——行已在，UPDATE
+ *   挂接非新建，见 handleContentRow 态 1b）；
  * - skipped：内容行未产修订的跳过数（同快照 checksum 命中、跨快照无业务变化、locale
  *   拒收三类；attachment 命中不计——媒体域是 Task 4 主场）；
  * - unsupportedRecords/failedRecords：unsupported/error 事件分项数（**按事件计**非按
  *   插入行——重跑行幂等而计数如实；拆分只在分派点可知，Task 6 报告层无从重derive）。
  * issues：locale 不匹配、未知 sourceType 兜底——显式报告（§7.6 禁静默）。
  */
-export function createContentEventHandler(identity: {
-  db: ImportDb
-  siteId: string
-  /** 快照行六列身份的 connection 侧 + 五列内容身份的解析锚点 */
-  connectionId: string
+export function createContentEventHandler(identity: WriterContext & {
   /** 本 job 的 wxr 快照 id（六列身份的 snapshot 侧 + 新行 source_snapshot_id） */
   snapshotId: string
   /** 本 job id（source_records.import_job_id 与 content_revisions.import_job_id 落库锚点） */
@@ -286,8 +283,6 @@ export function createContentEventHandler(identity: {
 
     // 态 1：本快照行已存在（挂 entry——本写入器是 content 行 source_records 的唯一写入方，
     // 挂接不变量由写入路径自身保证）。
-    // 已知死角：NULL-entry 占位行（外源 seed/未来写入方）+ 后续映射扩充重导 → 全新建分支
-    // 插入撞六列唯一约束 23505；Task 6 收编（UPDATE 该行挂 entry）——见审查记录
     const snapRow = await findSnapshotRecord(row)
     if (snapRow !== null && snapRow.contentEntryId !== null) {
       if (snapRow.normalizedChecksum === checksum) {
@@ -315,6 +310,67 @@ export function createContentEventHandler(identity: {
       counts.revisionsCreated++
       return
     }
+    if (snapRow !== null) {
+      // 态 1b：NULL-entry 占位行收编（Task 6 义务①——Task 3 审查移交的已知死角在此闭合）：
+      // 六列命中而 content_entry_id NULL 的行（外源 seed / attachment record-only 足迹 +
+      // 映射扩充后该 sourceType 升格为内容类型重导）→ 走全新建分支会撞六列唯一约束
+      // 23505，改为就地 UPDATE 挂 entry。语义与态 2/3 对齐：
+      // - 无五列先行者 → 全新建变体（entry+revision+locale），占位行 UPDATE 挂接；
+      // - 先行者 checksum 同 → 只挂接（skipped——0 新修订）；
+      // - 先行者 checksum 异 → 挂接 + 新修订 + current 前移。
+      // counts：sourceRecords 不计（行已在，收编非新建）；lineage 留在占位行建行 job。
+      const prior = await findPriorLinkedRecord(row)
+      const adoptValues = { contentEntryId: null as string | null, normalizedChecksum: checksum, sourceUrl: row.sourceUrl ?? null }
+      if (prior === null) {
+        await db.transaction(async (tx) => {
+          const entry = await tx
+            .insert(contentEntries)
+            .values({ siteId, typeKey, lifecycleStatus: 'active' })
+            .returning({ id: contentEntries.id })
+          const entryId = entry[0]!.id
+          await tx
+            .update(sourceRecords)
+            .set({ ...adoptValues, contentEntryId: entryId })
+            .where(eq(sourceRecords.id, snapRow.id))
+          const revisionId = await insertRevision(tx, {
+            entryId,
+            row,
+            data,
+            checksum,
+            sourceRecordId: snapRow.id,
+          })
+          await pointCurrent(tx, entryId, row, revisionId)
+        })
+        counts.contentEntries++
+        counts.revisionsCreated++
+        return
+      }
+      const entryId = prior.contentEntryId // 事务回调外的 const 收窄（属性收窄不进闭包）
+      if (prior.normalizedChecksum === checksum) {
+        await db
+          .update(sourceRecords)
+          .set({ ...adoptValues, contentEntryId: entryId })
+          .where(eq(sourceRecords.id, snapRow.id))
+        counts.skipped++
+        return
+      }
+      await db.transaction(async (tx) => {
+        await tx
+          .update(sourceRecords)
+          .set({ ...adoptValues, contentEntryId: entryId })
+          .where(eq(sourceRecords.id, snapRow.id))
+        const revisionId = await insertRevision(tx, {
+          entryId,
+          row,
+          data,
+          checksum,
+          sourceRecordId: snapRow.id,
+        })
+        await pointCurrent(tx, entryId, row, revisionId)
+      })
+      counts.revisionsCreated++
+      return
+    }
 
     // 态 2/3 前置：内容身份（五列）解析——跨快照复用既有 entry（permalink/快照变化不建新）
     const prior = await findPriorLinkedRecord(row)
diff --git a/apps/api/src/modules/import/import.service.ts b/apps/api/src/modules/import/import.service.ts
index d82ba45..9fa08de 100644
--- a/apps/api/src/modules/import/import.service.ts
+++ b/apps/api/src/modules/import/import.service.ts
@@ -1,4 +1,4 @@
-// import_jobs 生命周期服务（Plan 3 Task 1，spec §7.2/§7.5、前置事实 6/13）：
+// import_jobs 生命周期服务（Plan 3 Task 1 + Task 6 编排，spec §7.2/§7.5、前置事实 6/8/13）：
 // 解析站点与快照 → 幂等建 job → 状态推进（queued → running → succeeded /
 // succeeded_with_issues / failed）→ 报告落 <reportRoot>/<jobId>/reports/import.json
 // （temp+rename 原子写、**先于终态落行**——行内 report_path 出现即文件存在）。
@@ -21,24 +21,37 @@
 // site + wxr 快照 + 三语义版本（adapter/mapping/sanitizer——任一 bump 即新 job，对齐
 // spec §7.5 “版本变化可显式触发重新标准化”；parser 是诊断信息不进键）。
 //
-// 管道注入：Task 6 才接真实管道（terms/content/media/routes 写入器）；本任务默认**空
-// 管道**（零计数零问题直接成功），pipeline 钩子供集成测试观察 running 中间态、
-// succeeded_with_issues 与 failed 路径。
+// 管道注入：Task 6 起默认 = **编排管道**（四家写入器汇流，orchestratedPipeline）——真
+// 事件流（runWxrAdapter 读 wxr 快照副本，mapConfig.locale 接 site.default_locale——前置
+// 事实 8）→ content/term/media/route 严格串行分派 → finish 硬序（media 先于 route）→
+// counts 按键**求和**聚合（Task 6 钉死，见 ImportPipelineOutcome.counts 契约）。注入
+// pipeline 钩子供集成测试观察中间态/失败路径（注入时缺省事件流为空——Task 1 语义保持）。
+//
+// stale 回收（Task 6 义务④）：每次建 job 前把同 site 超过 24h 仍 queued/running 的 job
+// 置 stale（崩溃/中断残留终态化）。心跳列决策留 ADR-017——现阶段无心跳列，以
+// created_at 近似（job 为单进程就地推进，24h 远超任何真实执行时长）。
+import { createReadStream } from 'node:fs'
 import { mkdir, rename, rm, writeFile } from 'node:fs/promises'
 import { dirname, join, resolve } from 'node:path'
 import { fileURLToPath } from 'node:url'
-import { and, desc, eq } from 'drizzle-orm'
+import { and, desc, eq, inArray, lt } from 'drizzle-orm'
 import type { NodePgDatabase } from 'drizzle-orm/node-postgres'
 import {
   ADAPTER_VERSION,
   MAPPING_VERSION,
   PARSER_VERSION,
   SANITIZER_VERSION,
+  runWxrAdapter,
   type AdapterEvent,
+  type AdapterTimings,
 } from '@sf/adapter-wp'
 import { importJobs, sites, sourceConnections, sourceSnapshots } from '../../db/schema.js'
 import type * as schema from '../../db/schema.js'
 import { isUniqueViolation } from '../source/register.service.js'
+import { createContentEventHandler } from './content-writer.js'
+import { createTermEventHandler } from './term-writer.js'
+import { createMediaEventHandlerFromContext } from './media-writer.js'
+import { DEFAULT_NEWS_PATH, DEFAULT_PRODUCTS_PATH, createRouteEventHandler, normalizeListPath } from './route-writer.js'
 
 /** 本文件位于 <repo>/apps/api/src/modules/import/ → 上五级是仓库根 */
 const REPO_ROOT = resolve(dirname(fileURLToPath(import.meta.url)), '../../../../..')
@@ -109,37 +122,66 @@ export interface ImportJobContext {
   report: ImportPipelineOutcome
   /**
    * 适配器事件流（runWxrAdapter 产物：row/unsupported/error/term 按文档序交错）。
-   * Task 2 加法扩展：测试经 {@link CreateImportJobInput.events} 手构注入；Task 6 编排
-   * 时改为从 wxr 快照构造真流（runWxrAdapter(storagePath 流)）——多个写入器共享同流
-   * 时的重迭代策略（每写入器一新生成器等）由 Task 6 决定，本字段只承诺"可迭代一次"。
+   * Task 6 兑现：缺省 = 从 wxr 快照副本构造的真流（mapConfig.locale=site.default_locale
+   * ——前置事实 8）；测试经 {@link CreateImportJobInput.events} 手构注入。单消费方
+   * （编排器）一遍 for-await 分派给四家写入器——只承诺"可迭代一次"，无 tee 无重迭代。
    */
   events: AsyncIterable<AdapterEvent>
 }
 
 /**
- * 管道产出契约（Task 1 钉死，Task 6 写入器遵从）：
+ * 管道产出契约（Task 1 钉死、Task 6 修订——counts 聚合改**求和**、加法扩 issueGroups/
+ * sanitize）：
  * - counts：计数块（键名由写入器约定，如 contentEntries/terms/routes）——原样进报告；
  * - issues：问题清单（简单字符串——机器可读错误码按 Plan 2 台账推迟）；**非空 → job
- *   终态 succeeded_with_issues**（问题全部入报告，不静默）。
+ *   终态 succeeded_with_issues**（问题全部入报告，不静默）；
+ * - issueGroups / sanitize：可选分组视图与清洗统计（编排管道产出；issues 仍是全量扁平清单）。
  *
- * 汇合规则（钉死）：最终产出 = ctx.report 累加器 ⊕ 管道返回值——counts 按键合并
- * （**返回值覆盖累加器同名键**）、issues 串联（累加器在前、返回值在后）。管道抛错时
- * 无返回值，累加器现值原样进报告（部分产出可审计）。
+ * 汇合规则（Task 6 钉死）：最终产出 = ctx.report 累加器 ⊕ 管道返回值——counts 按键
+ * **求和**、issues 串联（累加器在前、返回值在后）、issueGroups 逐键串联、sanitize
+ * 数值相加。管道抛错时无返回值，累加器现值原样进报告（部分产出可审计）。
  */
 export interface ImportPipelineOutcome {
   /**
    * 计数块（原样进报告）。**键契约：counts 键是跨写入器的扁平全局命名空间**——
-   * 汇合规则按"返回值覆盖累加器同名键"合并（ctx.report.counts ⊕ outcome.counts），
-   * 同名键 **CLOBBER 不求和**：各写入器必须各持独占键名并在自身 JSDoc 声明归属
-   * （Task 2 term-writer：termsProcessed/autoCreatedTerms/contentTermsCreated；
-   * Task 3 content-writer：contentEntries/revisionsCreated/sourceRecords/skipped/
-   * unsupportedRecords/failedRecords——Task 4 media-writer 起新键名，勿复用 'skipped'）。
+   * 汇合按"同名键**求和**"聚合（ctx.report.counts ⊕ outcome.counts，Task 6 钉死：
+   * 四家写入器键前缀互斥，求和是防覆盖的安全网——Task 1 的 CLOBBER 语义废弃）。
+   * 各写入器仍须各持独占键名并在自身 JSDoc 声明归属（term：termsProcessed/
+   * autoCreatedTerms/contentTermsCreated；content：contentEntries/revisionsCreated/
+   * sourceRecords/skipped/unsupportedRecords/failedRecords；media 六键 media 前缀；
+   * route 五键 routes 前缀——新写入器勿复用他家键名）。
    */
   counts: Record<string, number>
   issues: string[]
+  /**
+   * 可选：按来源分组的 issues 视图（报告分组呈现）。**issues 字段仍是无分组全量
+   * 清单**（既有消费方契约不变）；分组键 = 四写入器前缀 + adapter（row 事件自带行级
+   * 问题）+ orchestrator（编排器自身）。
+   */
+  issueGroups?: Partial<Record<ImportIssueGroup, string[]>>
+  /** 可选：清洗统计（row 事件 sanitizeDiffs/changed 的聚合——编排层在分派循环收集） */
+  sanitize?: ImportSanitizeStats
 }
 
-/** 管道（Task 6 接真实写入器；缺省空管道=零计数零问题直接成功）。抛错 → job failed */
+/** issues 分组键：四写入器前缀 + adapter（row 事件 issues 面）+ orchestrator（编排器） */
+export type ImportIssueGroup = 'content' | 'term' | 'media' | 'route' | 'adapter' | 'orchestrator'
+
+/** 清洗统计（Plan 2 报告纪律的导入侧镜像：changed 语义 = html !== 输入原文） */
+export interface ImportSanitizeStats {
+  /** changed=true 的 row 数（零 diff 等价变形也计——如 img 自闭合规范化） */
+  changedRows: number
+  /** 按 diff 类型计数（'removed-tag'/'removed-attr'——sparse：零计类型不占键） */
+  byDiffType: Record<string, number>
+}
+
+/**
+ * 写入器共享身份基类型（Task 6 义务③：编译期防漂移）——四家写入器的 identity 参数均以
+ * 此为基础（`WriterContext & { 各家专属列 }`），db/siteId/connectionId 三列的形状与命名
+ * 单源约束；ImportJobContext 结构性满足（Pick 自它），编排注入即传 ctx 的这三列。
+ */
+export type WriterContext = Pick<ImportJobContext, 'db' | 'siteId' | 'connectionId'>
+
+/** 管道（缺省 = Task 6 编排管道 orchestratedPipeline——四家写入器汇流）。抛错 → job failed */
 export type ImportPipeline = (ctx: ImportJobContext) => Promise<ImportPipelineOutcome>
 
 export interface CreateImportJobInput {
@@ -153,13 +195,17 @@ export interface CreateImportJobInput {
   /** 报告根（落 <reportRoot>/<jobId>/reports/import.json）；缺省 repoRoot/artifacts/import */
   reportRoot?: string
   /**
-   * 适配器事件流（ctx.events 的注入位）：Task 2 测试手构注入；缺省空流（空管道下无消费
-   * 方）。Task 6 编排时由此改为从 wxr 快照构造 runWxrAdapter 真流（前置事实 8：mapConfig
-   * .locale 必须传 site.default_locale）。
+   * 适配器事件流（ctx.events 的注入位）：测试手构注入。注入而缺省 pipeline 时仍走编排
+   * 管道（手构流 + 真编排）；注入 pipeline 而无 events 时为空流（Task 1 语义保持）；
+   * 两者皆缺席 = 真流（wxr 快照副本，前置事实 8 的 locale 接线在真流构造处）。
    */
   events?: AsyncIterable<AdapterEvent>
-  /** 管道钩子（测试/Task 6 注入）；缺省空管道 */
+  /** 管道钩子（测试注入）；缺省 = 编排管道 */
   pipeline?: ImportPipeline
+  /** 主列表页路径（前置事实 9：CLI 参数来源，规范化值记入 job 报告）；缺省 /products/ */
+  productsPath?: string
+  /** 新闻列表页路径（同上）；缺省 /news/ */
+  newsPath?: string
 }
 
 export interface CreateImportJobResult {
@@ -279,6 +325,20 @@ export async function createImportJob(
   }
   const site = siteRows[0]!
 
+  // 阶段 0.5：stale 回收（Task 6 义务④）——同 site 超过 24h 仍 queued/running 的 job 置
+  // stale（崩溃/中断残留的终态化，finished_at 记宣判时刻）。心跳列决策留 ADR-017：现阶
+  // 段无心跳列，以 created_at 近似（job 单进程就地推进，24h 远超任何真实执行时长）。
+  await db
+    .update(importJobs)
+    .set({ status: 'stale', finishedAt: new Date() })
+    .where(
+      and(
+        eq(importJobs.siteId, site.id),
+        inArray(importJobs.status, ['queued', 'running']),
+        lt(importJobs.createdAt, new Date(Date.now() - STALE_AFTER_MS)),
+      ),
+    )
+
   // 阶段 1：wxr 主输入快照解析（显式 id 优先；缺省按 (connection, kind='wxr') 最新）
   let wxrRow: SnapshotRow | null = null
   if (input.snapshotId !== undefined) {
@@ -370,8 +430,23 @@ export async function createImportJob(
     .set({ status: 'running', startedAt: new Date(), attempt })
     .where(eq(importJobs.id, jobId))
 
-  // 阶段 5：管道（本任务缺省空管道；Task 6 接真实写入器）——抛错 → failed 终态。
-  // 产出汇合规则（ImportPipeline JSDoc 钉死）：返回值 counts 覆盖累加器同名键、issues 追加在后。
+  // 阶段 5：管道——缺省 = 编排管道（四家写入器汇流）；注入钩子优先。抛错 → failed 终态。
+  // 事件流三分：注入 events 优先（手构流 + 缺省编排管道 = 测试形态）→ 注入 pipeline 而
+  // 无 events = 空流（Task 1 语义保持）→ 双缺席 = 真流（wxr 快照副本 + locale 接线，
+  // 前置事实 8——缺省空串会让全部行被 content-writer 以 locale 不匹配拒收）。
+  const productsPath = normalizeListPath(input.productsPath ?? DEFAULT_PRODUCTS_PATH)
+  const newsPath = normalizeListPath(input.newsPath ?? DEFAULT_NEWS_PATH)
+  const useRealStream = input.events === undefined && input.pipeline === undefined
+  const adapterTimings: AdapterTimings | null = useRealStream ? { parseMs: 0, sanitizeMs: 0 } : null
+  const events: AsyncIterable<AdapterEvent> =
+    input.events !== undefined
+      ? input.events
+      : input.pipeline !== undefined
+        ? emptyEvents()
+        : realEventStream(wxrSnapshot, site.defaultLocale, adapterTimings!)
+  const pipeline: ImportPipeline =
+    input.pipeline !== undefined ? input.pipeline : orchestratedPipeline({ productsPath, newsPath })
+
   let failed: { code: ImportErrorCode; summary: string } | null = null
   const ctx: ImportJobContext = {
     db,
@@ -385,11 +460,11 @@ export async function createImportJob(
     manualUrlListSnapshot,
     versions: VERSIONS,
     report: { counts: {}, issues: [] },
-    events: input.events ?? emptyEvents(),
+    events,
   }
   let outcome: ImportPipelineOutcome | null = null
   try {
-    outcome = await (input.pipeline ?? emptyPipeline)(ctx)
+    outcome = await pipeline(ctx)
   } catch (err) {
     failed = {
       code: err instanceof ImportError ? err.code : 'PIPELINE_FAILED',
@@ -399,10 +474,7 @@ export async function createImportJob(
     }
   }
   // 抛错时无返回值：累加器现值进报告（部分产出可审计）
-  const merged: ImportPipelineOutcome =
-    outcome === null
-      ? ctx.report
-      : { counts: { ...ctx.report.counts, ...outcome.counts }, issues: [...ctx.report.issues, ...outcome.issues] }
+  const merged: ImportPipelineOutcome = outcome === null ? ctx.report : mergeOutcomes(ctx.report, outcome)
   let finalStatus: ImportJobStatus
   if (failed !== null) {
     finalStatus = 'failed'
@@ -431,8 +503,14 @@ export async function createImportJob(
       manual_url_list: manualUrlListSnapshot === null ? null : slimRef(manualUrlListSnapshot),
     },
     versions: VERSIONS,
+    // 主列表路径（前置事实 9：CLI 参数来源记入报告——route-writer 实际消费的规范化值）
+    paths: { productsPath, newsPath },
     counts: merged.counts,
     issues: merged.issues,
+    ...(merged.issueGroups === undefined ? {} : { issueGroups: merged.issueGroups }),
+    ...(merged.sanitize === undefined ? {} : { sanitize: merged.sanitize }),
+    // 适配器分段计时（仅真流编排——Plan 2 报告纪律的 parse/sanitize 分段在导入侧续录）
+    ...(adapterTimings === null ? {} : { adapterTiming: { parseMs: adapterTimings.parseMs, sanitizeMs: adapterTimings.sanitizeMs } }),
     error: failed,
     durationMs: performance.now() - startedAt,
   }
@@ -522,11 +600,163 @@ function slimRef(ref: SnapshotRef): {
   return { id: ref.id, kind: ref.kind, sha256: ref.sha256, byteSize: ref.byteSize, storagePath: ref.storagePath }
 }
 
-/** Task 1 空管道：零计数零问题直接成功（Task 6 以真实写入器替换缺省值） */
-async function emptyPipeline(_ctx: ImportJobContext): Promise<ImportPipelineOutcome> {
-  return { counts: {}, issues: [] }
-}
-
 /** 缺省空事件流（input.events 缺席时的 ctx.events 占位；每次调用产新迭代器） */
 async function* emptyEvents(): AsyncGenerator<AdapterEvent, void, void> {}
 
+// ---- 编排管道（Task 6 汇流——Task 5 审查移交的契约底本在此兑现） ----
+
+/** stale 回收阈值：queued/running 滞留超过此时长即宣判（毫秒） */
+const STALE_AFTER_MS = 24 * 60 * 60 * 1000
+
+/** wxr 快照副本绝对路径（metadata.storageRoot + storagePath——登记期固化的存储根事实） */
+function wxrSnapshotCopyPath(snapshot: SnapshotRef): string {
+  const root = snapshot.metadata.storageRoot
+  if (typeof root !== 'string' || root === '') {
+    throw new Error(
+      `wxr 快照 ${snapshot.id} metadata 缺 storageRoot（登记期不变量破裂——无法读取副本构造事件流）`,
+    )
+  }
+  return join(root, snapshot.storagePath)
+}
+
+/**
+ * 真事件流（惰性生成器：路径解析/读盘错误推迟到首拍 next()——落进管道执行期，由
+ * createImportJob 的 PIPELINE_FAILED 收口，不留"job 已 running 而流构造抛错"的悬挂行）。
+ * 前置事实 8：mapConfig.locale 必须接 site.default_locale（缺省空串 = 全行被拒收）。
+ */
+async function* realEventStream(
+  snapshot: SnapshotRef,
+  locale: string,
+  timings: AdapterTimings,
+): AsyncGenerator<AdapterEvent, void, void> {
+  yield* runWxrAdapter(createReadStream(wxrSnapshotCopyPath(snapshot)), {
+    mapConfig: { locale },
+    timings,
+  })
+}
+
+/** counts 按键求和（Task 6 钉死的聚合策略——就地累加 target） */
+function sumCounts(target: Record<string, number>, source: Record<string, number>): void {
+  for (const [key, value] of Object.entries(source)) {
+    target[key] = (target[key] ?? 0) + value
+  }
+}
+
+/** Outcome 汇合（ImportPipelineOutcome JSDoc 契约）：counts 求和、issues/issueGroups 串联、sanitize 相加 */
+function mergeOutcomes(acc: ImportPipelineOutcome, ret: ImportPipelineOutcome): ImportPipelineOutcome {
+  const counts: Record<string, number> = { ...acc.counts }
+  sumCounts(counts, ret.counts)
+  const issueGroups: Partial<Record<ImportIssueGroup, string[]>> = {}
+  let hasGroups = false
+  for (const key of Object.keys(acc.issueGroups ?? {}) as ImportIssueGroup[]) {
+    issueGroups[key] = [...(acc.issueGroups![key] ?? [])]
+    hasGroups = true
+  }
+  for (const key of Object.keys(ret.issueGroups ?? {}) as ImportIssueGroup[]) {
+    issueGroups[key] = [...(issueGroups[key] ?? []), ...(ret.issueGroups![key] ?? [])]
+    hasGroups = true
+  }
+  const sanitize: ImportSanitizeStats | undefined =
+    acc.sanitize === undefined && ret.sanitize === undefined
+      ? undefined
+      : {
+          changedRows: (acc.sanitize?.changedRows ?? 0) + (ret.sanitize?.changedRows ?? 0),
+          byDiffType: {},
+        }
+  if (sanitize !== undefined) {
+    for (const src of [acc.sanitize?.byDiffType, ret.sanitize?.byDiffType]) {
+      if (src !== undefined) sumCounts(sanitize.byDiffType, src)
+    }
+  }
+  return {
+    counts,
+    issues: [...acc.issues, ...ret.issues],
+    ...(hasGroups ? { issueGroups } : {}),
+    ...(sanitize === undefined ? {} : { sanitize }),
+  }
+}
+
+/**
+ * 编排管道（Task 6 汇流契约底本——Task 5 审查移交钉死）：
+ * - 事件期**严格串行 await** 分派 content → term → media → route：三家（term/media/
+ *   route）的 entry 锚点解析依赖 content 对**同事件**的落库产物（source_records.
+ *   content_entry_id / entry_locales），禁 Promise.all——并行下锚点查询可能抢在 content
+ *   写入前发生；
+ * - finish 唯一硬序：term/content（同步、无库依赖）任意 → `await media.finish()` 先于
+ *   `await route.finish()`（attachment 301 反查与 asset 全表都读 media_assets——资产须已齐）；
+ * - 分派规则（Task 3 审查移交⑥）：wp:attachment 行的媒体域归 media-writer
+ *   （media_assets 唯一写入方），content-writer 对其只留 source_records record-only 足迹
+ *   （前置事实 11）——两家各写各域，单一 owner；
+ * - counts 按键求和聚合（安全网——四家键前缀互斥）；清洗统计与 row 事件行级问题在
+ *   分派循环收集（RowResult 契约：issues 逐条收录为行级错误）。
+ */
+function orchestratedPipeline(paths: { productsPath: string; newsPath: string }): ImportPipeline {
+  return async (ctx) => {
+    const content = createContentEventHandler({
+      db: ctx.db,
+      siteId: ctx.siteId,
+      connectionId: ctx.connectionId,
+      snapshotId: ctx.wxrSnapshot.id,
+      jobId: ctx.jobId,
+      defaultLocale: ctx.defaultLocale,
+      versions: ctx.versions,
+    })
+    const term = createTermEventHandler({
+      db: ctx.db,
+      siteId: ctx.siteId,
+      connectionId: ctx.connectionId,
+    })
+    const media = await createMediaEventHandlerFromContext(ctx)
+    const route = createRouteEventHandler({
+      db: ctx.db,
+      siteId: ctx.siteId,
+      connectionId: ctx.connectionId,
+      snapshotId: ctx.wxrSnapshot.id,
+      defaultLocale: ctx.defaultLocale,
+      productsPath: paths.productsPath,
+      newsPath: paths.newsPath,
+    })
+
+    const sanitize: ImportSanitizeStats = { changedRows: 0, byDiffType: {} }
+    const adapterIssues: string[] = []
+    for await (const ev of ctx.events) {
+      if (ev.kind === 'row') {
+        if (ev.changed) sanitize.changedRows++
+        for (const diff of ev.sanitizeDiffs) {
+          sanitize.byDiffType[diff.type] = (sanitize.byDiffType[diff.type] ?? 0) + 1
+        }
+        for (const issue of ev.issues) {
+          adapterIssues.push(`行级问题（${ev.row.sourceType}/${ev.row.sourceId}）：${issue}`)
+        }
+      }
+      await content.handleEvent(ev)
+      await term.handleEvent(ev)
+      await media.handleEvent(ev)
+      await route.handleEvent(ev)
+    }
+    const t = term.finish()
+    const c = content.finish()
+    const m = await media.finish()
+    const r = await route.finish()
+
+    const counts: Record<string, number> = {}
+    sumCounts(counts, c.counts)
+    sumCounts(counts, t.counts)
+    sumCounts(counts, m.counts)
+    sumCounts(counts, r.counts)
+    return {
+      counts,
+      issues: [...c.issues, ...t.issues, ...m.issues, ...r.issues, ...adapterIssues],
+      issueGroups: {
+        content: c.issues,
+        term: t.issues,
+        media: m.issues,
+        route: r.issues,
+        adapter: adapterIssues,
+        orchestrator: [],
+      },
+      sanitize,
+    }
+  }
+}
+
diff --git a/apps/api/src/modules/import/media-writer.ts b/apps/api/src/modules/import/media-writer.ts
index c7c6c81..b5af482 100644
--- a/apps/api/src/modules/import/media-writer.ts
+++ b/apps/api/src/modules/import/media-writer.ts
@@ -43,7 +43,7 @@ import type { UniversalRow } from '@sf/contracts'
 import { entryLocales, mediaAssets, mediaUsages, sourceRecords } from '../../db/schema.js'
 import { isUniqueViolation, type MediaManifestEntry } from '../source/register.service.js'
 import { detectMediaType } from './magic-bytes.js'
-import type { ImportDb, ImportJobContext, ImportPipeline, ImportPipelineOutcome } from './import.service.js'
+import type { ImportJobContext, ImportPipeline, ImportPipelineOutcome, WriterContext } from './import.service.js'
 
 const UPLOADS_MARKER = '/wp-content/uploads/'
 
@@ -119,10 +119,7 @@ interface SidecarEntry {
   filePath: string
 }
 
-export function createMediaEventHandler(identity: {
-  db: ImportDb
-  siteId: string
-  connectionId: string
+export function createMediaEventHandler(identity: WriterContext & {
   snapshotId: string
   /** 引用行 locale 校验基准（与 content-writer 同拒收口径——不双报） */
   defaultLocale: string
diff --git a/apps/api/src/modules/import/route-writer.ts b/apps/api/src/modules/import/route-writer.ts
index cbfdf2c..6693dc1 100644
--- a/apps/api/src/modules/import/route-writer.ts
+++ b/apps/api/src/modules/import/route-writer.ts
@@ -72,7 +72,7 @@ import type { AdapterEvent } from '@sf/adapter-wp'
 import type { UniversalRow } from '@sf/contracts'
 import { mediaAssets, routeRegistry, sites, sourceRecords, taxonomies, terms } from '../../db/schema.js'
 import { CONTENT_TYPE_KEY_MAP } from './content-writer.js'
-import type { ImportDb, ImportPipeline, ImportPipelineOutcome } from './import.service.js'
+import type { ImportPipeline, ImportPipelineOutcome, WriterContext } from './import.service.js'
 
 /** CLI --products-path 默认值（前置事实 9：记入 job metadata，本写入器收规范化值） */
 export const DEFAULT_PRODUCTS_PATH = '/products/'
@@ -88,8 +88,8 @@ export function foldCollisionKey(path: string): string {
   return path.normalize('NFC').toLowerCase().replace(/\/+$/, '')
 }
 
-/** 主列表路径规范化（CLI 参数防线：补首尾斜杠） */
-function normalizeListPath(path: string): string {
+/** 主列表路径规范化（CLI 参数防线：补首尾斜杠）；Task 6 起编排层导入用于报告记录实际生效值 */
+export function normalizeListPath(path: string): string {
   const withLeading = path.startsWith('/') ? path : `/${path}`
   return withLeading.endsWith('/') ? withLeading : `${withLeading}/`
 }
@@ -176,11 +176,7 @@ interface RoutePayload {
  * identity 注记：connectionId/snapshotId 是 entry 锚点解析与 source_snapshot_id 溯源的
  * 必需输入（与 content/media 同源——Task 6 从 ImportJobContext 共享注入）。
  */
-export function createRouteEventHandler(identity: {
-  db: ImportDb
-  siteId: string
-  /** source_records 五列身份解析的连接锚点（content-writer 同口径） */
-  connectionId: string
+export function createRouteEventHandler(identity: WriterContext & {
   /** 行溯源（route_registry.source_snapshot_id） */
   snapshotId: string
   /** locale 校验基准（与 content/media 同拒收口径——不双报） */
@@ -443,20 +439,27 @@ export function createRouteEventHandler(identity: {
 
   /**
    * finish 期单条 attachment 处理：URL 层防线（link 在场 → 解析 → ?p=/&page_id= 排除 → 路径
-   * 安全）先行，资产反查在后 → 301 载荷（查询/permalink 双形态）。返回 true 仅当实际写入
-   * （created/updated——重复事件 no-op 不重复计入降级汇总）。
+   * 安全）先行，资产反查在后 → 301 载荷（查询/permalink 双形态）。返回三态：
+   * - 'written'：实际写入（created/updated）；
+   * - 'resolved'：目标解析成功且行全等 no-op（幂等重跑——降级处置照常成立，只是零写入）；
+   * - 'blocked'：目标缺席/歧义/防线拒绝（issue 已记，路由未登记）。
+   * 降级汇总按 **去重 source_id 的 resolved∪written** 计（Task 6 口径：降级是处置事实，
+   * 重跑如实再现；同 attachment 重复事件不虚报——Plan 3 钉死语义）。
    */
-  async function writeAttachmentRoute(att: PendingAttachment, host: string): Promise<boolean> {
+  async function writeAttachmentRoute(
+    att: PendingAttachment,
+    host: string,
+  ): Promise<'written' | 'resolved' | 'blocked'> {
     if (att.sourceUrl === '') {
       issues.push(`attachment 缺 link（wp:attachment/${att.sourceId}，${att.rawLocation}）：无页面 URL 证据——301 路由不登记`)
-      return false
+      return 'blocked'
     }
     let parsed: URL
     try {
       parsed = new URL(att.sourceUrl)
     } catch (err) {
       issues.push(`attachment link 解析失败（wp:attachment/${att.sourceId}，${att.sourceUrl}）：${err instanceof Error ? err.message : String(err)}——301 路由不登记`)
-      return false
+      return 'blocked'
     }
     // ?p=/&page_id= 排除口径与内容行一致（Fix 对齐：attachment link 同形态同排除——
     // excluded_with_reason 语义见模块头；?attachment_id= 不在排除表，照常 query_key 建模）
@@ -464,12 +467,12 @@ export function createRouteEventHandler(identity: {
       if (parsed.searchParams.has(param)) {
         counts.routesExcludedQueries++
         issues.push(`查询 URL 排除（wp:attachment/${att.sourceId}，${att.sourceUrl}）：?${param}= 形态不登记 route_registry（route_kind 无 excluded 值——查询路由 SP3；spec §6.1/§6.2 excluded_with_reason 以计数+本 issue 落报告）`)
-        return false
+        return 'blocked'
       }
     }
     if (!isSafePath(parsed.pathname)) {
       issues.push(`attachment 路径安全校验拒绝（wp:attachment/${att.sourceId}，${parsed.pathname}）：含 NUL/反斜杠/控制字符/穿越段——301 路由不登记（完整校验规则 SP3 验证器）`)
-      return false
+      return 'blocked'
     }
     const assetRows = await db
       .select({ publicPath: mediaAssets.publicPath })
@@ -485,12 +488,12 @@ export function createRouteEventHandler(identity: {
       .limit(2)
     if (assetRows.length === 0) {
       issues.push(`attachment 301 目标缺席（wp:attachment/${att.sourceId}，${att.sourceUrl}）：media_assets 无对应资产（媒体三级对拍 miss？）——attachment 页路由未登记`)
-      return false
+      return 'blocked'
     }
     if (assetRows.length > 1) {
       // LIMIT 2 探针只知下界——歧义消息按 ≥2 表述（不自动选，spec §5.5）
       issues.push(`attachment 多资产歧义（wp:attachment/${att.sourceId}）：同 source_id 命中 ≥2 行——不自动选（spec §5.5），attachment 页路由未登记`)
-      return false
+      return 'blocked'
     }
     const publicPath = assetRows[0]!.publicPath
     const queryKey = canonicalQueryKey(parsed)
@@ -513,7 +516,7 @@ export function createRouteEventHandler(identity: {
       priority: 0,
       verificationStatus: 'verified_redirect',
     })
-    return outcome === 'created' || outcome === 'updated'
+    return outcome === 'created' || outcome === 'updated' ? 'written' : 'resolved'
   }
 
   return {
@@ -637,13 +640,15 @@ export function createRouteEventHandler(identity: {
         }
       }
 
-      // ③ 暂存 attachment 301——成功计数汇总为一条降级 issue（前置事实 10）
-      let downgraded = 0
+      // ③ 暂存 attachment 301——降级汇总为一条 issue（前置事实 10）；计数按去重
+      // source_id 的已解析处置数（written∪resolved）：幂等重跑 no-op 零写入但降级处置
+      // 照常成立、如实再现（Plan 3 钉死）；同 attachment 重复事件不虚报（去重）
+      const downgradedSources = new Set<string>()
       for (const att of pendingAttachments) {
-        if (await writeAttachmentRoute(att, host)) downgraded++
+        if ((await writeAttachmentRoute(att, host)) !== 'blocked') downgradedSources.add(att.sourceId)
       }
-      if (downgraded > 0) {
-        issues.push(`attachment 页降级（前置事实 10）：${downgraded} 条 attachment 页一律 301 → 媒体文件——仅 WXR 证据、无 sitemap 解析（sitemap→html 分支留待 SP3）`)
+      if (downgradedSources.size > 0) {
+        issues.push(`attachment 页降级（前置事实 10）：${downgradedSources.size} 条 attachment 页一律 301 → 媒体文件——仅 WXR 证据、无 sitemap 解析（sitemap→html 分支留待 SP3）`)
       }
 
       // ④ asset 行（media_assets 全表——site 的旧 URL 空间台账；source_url 是来源原文）
diff --git a/apps/api/src/modules/import/term-writer.ts b/apps/api/src/modules/import/term-writer.ts
index 8e9745d..cbb1303 100644
--- a/apps/api/src/modules/import/term-writer.ts
+++ b/apps/api/src/modules/import/term-writer.ts
@@ -36,7 +36,7 @@ import type { AdapterEvent, WxrTerm } from '@sf/adapter-wp'
 import type { UniversalRow } from '@sf/contracts'
 import { contentTerms, sourceRecords, taxonomies, terms } from '../../db/schema.js'
 import { isUniqueViolation } from '../source/register.service.js'
-import type { ImportDb, ImportPipeline, ImportPipelineOutcome } from './import.service.js'
+import type { ImportPipeline, ImportPipelineOutcome, WriterContext } from './import.service.js'
 
 /** 自建 term 的 source_id 前缀（与 wp:term 的 term_id 字符串源身份空间隔离） */
 const AUTO_SOURCE_ID_PREFIX = 'auto:'
@@ -94,12 +94,7 @@ export interface TermEventHandler {
  * issues：无 entry 可关联的 row 分类引用、表外 taxonomy 兜底、缺 term_id 的 wp:term
  * 事件、同 slug 双 term_id、收编撞唯一约束——全部显式报告（§7.6 禁静默）。
  */
-export function createTermEventHandler(identity: {
-  db: ImportDb
-  siteId: string
-  /** entry 解析（source_records 稳定身份查询）用的连接锚点 */
-  connectionId: string
-}): TermEventHandler {
+export function createTermEventHandler(identity: WriterContext): TermEventHandler {
   const db = identity.db
   const siteId = identity.siteId
   const counts = { termsProcessed: 0, autoCreatedTerms: 0, contentTermsCreated: 0 }
diff --git a/scripts/sf.ts b/scripts/sf.ts
index 7cbb904..d8e288b 100644
--- a/scripts/sf.ts
+++ b/scripts/sf.ts
@@ -2,7 +2,7 @@
 // sf CLI（Plan 2 Task 7 + Plan 3 Task 1）：commander 三命令——
 // - wxr:probe：只读诊断（任意可读路径；Task 8 直读 cms 用此通道），出 JSON/HTML 报告不入库；
 // - source:register：调 registerSource 登记 content-addressed 快照（路径白名单校验在服务内）；
-// - import：调 createImportJob 建 import_jobs 并推进生命周期（Task 1 空管道骨架；Task 6 接真实管道）。
+// - import：调 createImportJob 建 import_jobs 并推进生命周期（Task 6：默认编排管道——四家写入器汇流）。
 // 执行机制（Plan 2 评审钉死）：根 devDep tsx 直吃 TS 源码（含 apps/api 的 Nest 源码与
 // packages 的装饰器/TS 源文件），@sf/* 经根/包 tsconfig paths 直连 src——无需预构建 dist。
 // 注意：CLI 只 import register.service.ts / import.service.ts（零装饰器纯函数文件），
@@ -163,16 +163,30 @@ program
 
 program
   .command('import')
-  .description('导入已登记的 wxr 快照（Task 1 空管道骨架：job 生命周期 + 报告；真实管道 Task 6 接入）')
+  .description('导入已登记的 wxr 快照（Task 6 编排管道：term/content/media/route 四家写入器汇流 + 导入报告）')
   .requiredOption('--site <key>', '站点 key')
   .option('--snapshot <id>', 'wxr 快照 id（缺省按 connection+kind=wxr 选最新）')
   .option('--key <key>', '幂等键（缺省 = site+snapshot+版本块 自动派生；重复提交同 key 复用既有 job 不重跑）')
   .option('--connection <name>', '连接名（快照解析的 connection 侧，缺省 default）', 'default')
-  .action(async (opts: { site: string; snapshot?: string; key?: string; connection: string }) => {
+  .option('--products-path <path>', '产品主列表页路径（前置事实 9，规范化值记入 job 报告）', '/products/')
+  .option('--news-path <path>', '新闻主列表页路径（同上）', '/news/')
+  .action(async (opts: {
+    site: string
+    snapshot?: string
+    key?: string
+    connection: string
+    productsPath: string
+    newsPath: string
+  }) => {
     // getDb 走 DATABASE_URL_TEST ?? DATABASE_URL 回退链——临时库冒烟只需 env 覆盖
     const db = getDb()
     // exactOptionalPropertyTypes：缺席键不传 undefined 值（与 source:register 同款）
-    const input: CreateImportJobInput = { siteKey: opts.site, connectionName: opts.connection }
+    const input: CreateImportJobInput = {
+      siteKey: opts.site,
+      connectionName: opts.connection,
+      productsPath: opts.productsPath,
+      newsPath: opts.newsPath,
+    }
     if (opts.snapshot !== undefined) input.snapshotId = opts.snapshot
     if (opts.key !== undefined) input.idempotencyKey = opts.key
     try {
diff --git a/tests/integration/import-full.test.ts b/tests/integration/import-full.test.ts
new file mode 100644
index 0000000..5341bcf
--- /dev/null
+++ b/tests/integration/import-full.test.ts
@@ -0,0 +1,470 @@
+// Plan 3 Task 6 失败测试（先于实现）：编排集成 + 导入报告——四家写入器汇流。
+// 覆盖契约（对照计划 Task 6 与前置事实 8/11/13、Task 3/5 审查移交清单）：
+// 1. 全链（register wxr+media+media-map → 默认编排管道·真事件流）：content_entries=3 /
+//    source_records=5 / terms=2+自建 0 / content_terms=1 / media_assets=2 / routes=12；
+//    published_revision_id 全库 NULL（导入不发布——ADR-010）；mapConfig.locale 接线
+//    （site.default_locale='zh-CN' → 全部 source_records.locale='zh-CN'——locale 只可能
+//    来自编排接线，mini 无自带 locale）；报告 = 四家 counts 18 键全量 + 清洗统计 +
+//    版本块 + issues 分组 + 路径块 + 适配器分段计时；
+// 2. 二导幂等（换 key 重跑）：全部 counts 归零语义（termsProcessed/mediaDanglingRefs 等
+//    按事件计键如实再现）、数据库行数不变、job 仍 succeeded_with_issues
+//    （attachment 降级汇总 issue 必在——Plan 3 钉死语义：降级是处置事实非写入计数）；
+// 3. 无 media 快照失败路径：media 域全 missing 报告不失败（空 manifest 语义已备）；
+// 4. NULL-entry 占位行收编（Task 3 审查移交①）：六列命中 content_entry_id NULL 行 →
+//    UPDATE 挂 entry（不撞 23505、sourceRecords 不虚增）；
+// 5. stale 回收（义务④）：同 site 超 24h 的 queued/running 置 stale，新近 job 不受影响；
+// 6. REPORT_WRITE_FAILED（Task 1 遗留）：报告写盘失败 → failed 终态、report_path 不落。
+// 运行于 globalSetup 临时库（ADR-015）；storage/报告根 os.tmpdir，finally 清理。
+import { mkdir, mkdtemp, readFile, rm, writeFile } from 'node:fs/promises'
+import { tmpdir } from 'node:os'
+import { join, resolve } from 'node:path'
+import { fileURLToPath } from 'node:url'
+import { describe, expect, it } from 'vitest'
+import { createImportJob } from '../../apps/api/src/modules/import/import.service.js'
+import { registerSource } from '../../apps/api/src/modules/source/register.service.js'
+import { rand, withImmutableCleanup, withTestDb, type Db } from './helpers.js'
+
+const FIXTURES_ROOT = fileURLToPath(new URL('../fixtures', import.meta.url))
+const WXR_FIXTURE = resolve(FIXTURES_ROOT, 'wxr', 'mini.xml')
+
+const PNG_SIG = Buffer.from([0x89, 0x50, 0x4e, 0x47, 0x0d, 0x0a, 0x1a, 0x0a])
+
+/** 伪 PNG 体（魔数校验只看签名——载荷唯一避免跨测试撞哈希） */
+function pngBytes(seed: string): Buffer {
+  return Buffer.concat([PNG_SIG, Buffer.from(`png-body:${seed}`, 'latin1')])
+}
+
+/** 本文件写入的行清场：FK 反序（与 import-routes 同清单；两张不可变表走触发器开关） */
+async function cleanupSiteRows(db: Db, siteId: string): Promise<void> {
+  await db.execute(`delete from route_registry where site_id = '${siteId}'`)
+  await db.execute(`delete from media_usages where site_id = '${siteId}'`)
+  await db.execute(`delete from media_assets where site_id = '${siteId}'`)
+  await db.execute(`delete from content_terms where site_id = '${siteId}'`)
+  await db.execute(`delete from terms where site_id = '${siteId}'`)
+  await db.execute(`delete from taxonomies where site_id = '${siteId}'`)
+  await db.execute(`delete from source_records where site_id = '${siteId}'`)
+  await db.execute(`delete from entry_locales where site_id = '${siteId}'`)
+  await withImmutableCleanup(db, 'content_revisions', 'content_revisions_immutable', `site_id = '${siteId}'`)
+  await db.execute(`delete from content_entries where site_id = '${siteId}'`)
+  await db.execute(`delete from import_jobs where site_id = '${siteId}'`)
+  await withImmutableCleanup(db, 'source_snapshots', 'source_snapshots_immutable', `site_id = '${siteId}'`)
+  await db.execute(`delete from source_connections where site_id = '${siteId}'`)
+  await db.execute(`delete from sites where id = '${siteId}'`)
+}
+
+interface Work {
+  work: string
+  storageRoot: string
+  reportRoot: string
+  allowedRoots: string[]
+}
+
+async function makeWork(prefix: string): Promise<Work> {
+  const work = await mkdtemp(join(tmpdir(), `${prefix}-`))
+  return {
+    work,
+    storageRoot: join(work, 'storage'),
+    reportRoot: join(work, 'artifacts', 'import'),
+    allowedRoots: [FIXTURES_ROOT, work],
+  }
+}
+
+/** 全量登记（wxr + 媒体目录 + 侧车）——形态同 import-routes 全链用例；locale 可选注入 */
+async function registerFull(
+  db: Db,
+  w: Work,
+  spec: { locale?: string } = {},
+): Promise<{ key: string; siteId: string; wxrId: string }> {
+  const mediaDir = join(w.work, 'media')
+  await mkdir(join(mediaDir, '2026/10'), { recursive: true })
+  await writeFile(join(mediaDir, '2026/10/fixture-a.png'), pngBytes(`full-a-${rand()}`))
+  const guidB = `b-${rand()}.png`
+  await writeFile(join(mediaDir, guidB), pngBytes(`full-b-${rand()}`))
+  const sidecarPath = join(w.work, 'media-map.json')
+  await writeFile(
+    sidecarPath,
+    JSON.stringify([
+      { sourceUrl: 'https://fixture.example/wp-content/uploads/2026/10/fixture-b.png', filePath: guidB },
+    ]),
+  )
+  const key = `sffull-${rand()}`
+  const reg = await registerSource(db, {
+    siteKey: key,
+    wxrPath: WXR_FIXTURE,
+    mediaDir,
+    mediaMapPath: sidecarPath,
+    allowedRoots: w.allowedRoots,
+    storageRoot: w.storageRoot,
+    reportRoot: join(w.work, 'artifacts', 'register'),
+    ...(spec.locale === undefined ? {} : { siteDefaults: { defaultLocale: spec.locale } }),
+  })
+  return { key, siteId: reg.siteId, wxrId: reg.snapshots.find((s) => s.kind === 'wxr')!.id }
+}
+
+/** 仅 wxr 登记（失败路径/收编场景——media_bundle/manual_url_list 缺席） */
+async function registerWxrOnly(db: Db, w: Work): Promise<{ key: string; siteId: string; wxrId: string }> {
+  const key = `sffull-${rand()}`
+  const reg = await registerSource(db, {
+    siteKey: key,
+    wxrPath: WXR_FIXTURE,
+    allowedRoots: w.allowedRoots,
+    storageRoot: w.storageRoot,
+    reportRoot: join(w.work, 'artifacts', 'register'),
+  })
+  return { key, siteId: reg.siteId, wxrId: reg.snapshots.find((s) => s.kind === 'wxr')!.id }
+}
+
+/** 导入报告读取（默认编排管道的产物断言面） */
+async function readReport(w: Work, jobId: string): Promise<any> {
+  return JSON.parse(await readFile(join(w.reportRoot, jobId, 'reports', 'import.json'), 'utf8'))
+}
+
+/** 单值计数便捷 */
+async function countOf(db: Db, sql: string): Promise<number> {
+  const rows = (await db.execute(sql)) as any
+  return Number(rows[0].n)
+}
+
+describe('编排集成 + 导入报告（四家写入器汇流）', () => {
+  it('全链（真事件流 + 默认编排管道）：18 键 counts 全量、published 全 NULL、locale 接线、清洗统计、issues 分组、报告路径/版本/计时块', async () => {
+    await withTestDb(async (db) => {
+      const w = await makeWork('sf-full-chain')
+      try {
+        // locale 接线断言的实证设计（前置事实 8）：site 默认 locale 取 'zh-CN'——mini 无
+        // 自带 locale，行 locale 只可能来自编排接线 mapConfig.locale=site.default_locale
+        const { key, siteId, wxrId } = await registerFull(db, w, { locale: 'zh-CN' })
+        try {
+          const result = await createImportJob(db, {
+            siteKey: key,
+            snapshotId: wxrId,
+            reportRoot: w.reportRoot,
+          })
+          expect(result.reused).toBe(false)
+          // 悬空引用（media 203/204）×2 + attachment 降级汇总 ×1 → issues 非空
+          expect(result.status).toBe('succeeded_with_issues')
+
+          // 数据库行数对拍（计划 Task 6 钉死清单）
+          expect(await countOf(db, `select count(*)::int as n from content_entries where site_id = '${siteId}'`)).toBe(3)
+          expect(await countOf(db, `select count(*)::int as n from source_records where site_id = '${siteId}'`)).toBe(5)
+          expect(await countOf(db, `select count(*)::int as n from terms where site_id = '${siteId}'`)).toBe(2)
+          expect(await countOf(db, `select count(*)::int as n from content_terms where site_id = '${siteId}'`)).toBe(1)
+          expect(await countOf(db, `select count(*)::int as n from media_assets where site_id = '${siteId}'`)).toBe(2)
+          expect(await countOf(db, `select count(*)::int as n from route_registry where site_id = '${siteId}'`)).toBe(12)
+
+          // published 全库 NULL（导入不发布——ADR-010）+ current 全设（导入产出工作修订）
+          expect(
+            await countOf(db, `select count(*)::int as n from entry_locales where site_id = '${siteId}' and published_revision_id is not null`),
+          ).toBe(0)
+          expect(
+            await countOf(db, `select count(*)::int as n from entry_locales where site_id = '${siteId}' and current_revision_id is not null`),
+          ).toBe(3)
+
+          // locale 接线（前置事实 8）：全部 source_records 行 locale = site 默认（非空串）
+          const locales = (await db.execute(`select distinct locale from source_records where site_id = '${siteId}'`)) as any
+          expect(locales.map((r: any) => r.locale).sort()).toStrictEqual(['zh-CN'])
+
+          // attachment 分派规则（前置事实 11）：2 条 attachment 行是 record-only 足迹（无 entry）
+          expect(
+            await countOf(db, `select count(*)::int as n from source_records where site_id = '${siteId}' and source_type = 'wp:attachment' and content_entry_id is null`),
+          ).toBe(2)
+
+          // content_terms：post 101 的 category 引用（is_primary=true——WP 单主分类近似）
+          // （投影转 Boolean——postgres.js 原型差异会让 toStrictEqual 报"无视觉差异"失败）
+          const ct = (await db.execute(`select ct.is_primary, t.slug from content_terms ct
+            join terms t on t.id = ct.term_id where ct.site_id = '${siteId}'`)) as any
+          expect(ct.map((r: any) => ({ isPrimary: Boolean(r.is_primary), slug: r.slug }))).toStrictEqual([
+            { isPrimary: true, slug: 'company' },
+          ])
+
+          // 报告：四家 counts 18 键全量（全局命名空间求和聚合——键前缀互斥）
+          const report = await readReport(w, result.jobId)
+          expect(report.counts).toStrictEqual({
+            contentEntries: 3,
+            revisionsCreated: 3,
+            sourceRecords: 5,
+            skipped: 0,
+            termsProcessed: 2,
+            autoCreatedTerms: 0,
+            contentTermsCreated: 1,
+            mediaRegistered: 2,
+            mediaFromSidecar: 1,
+            mediaMissing: 0,
+            mediaQuarantined: 0,
+            mediaDanglingRefs: 2,
+            mediaAmbiguous: 0,
+            routesCreated: 12,
+            routesUpdated: 0,
+            routesRedirects: 2,
+            routesCollisions: 0,
+            routesExcludedQueries: 0,
+          })
+
+          // 清洗统计（分派循环收集 row 事件 sanitizeDiffs/changed）：mini 仅 post 101 正文
+          // img 自闭合规范化零 diff 变形（changedRows=1、无 diff 条目——sparse 形态）
+          expect(report.sanitize).toStrictEqual({ changedRows: 1, byDiffType: {} })
+
+          // issues 分组（按写入器前缀）：flat issues 保持全量（3 条），分组归位
+          expect(report.issues).toHaveLength(3)
+          expect(report.issueGroups).toStrictEqual({
+            content: [],
+            term: [],
+            media: report.issues.filter((i: string) => i.includes('悬空')),
+            route: report.issues.filter((i: string) => i.includes('attachment 页降级')),
+            adapter: [],
+            orchestrator: [],
+          })
+          expect(report.issueGroups.media).toHaveLength(2)
+          expect(report.issueGroups.route).toHaveLength(1)
+
+          // 路径块（前置事实 9：CLI 参数来源记入报告——缺省值）+ 版本块 + 分段计时（真流）
+          expect(report.paths).toStrictEqual({ productsPath: '/products/', newsPath: '/news/' })
+          expect(report.versions).toStrictEqual({
+            adapter: report.versions.adapter,
+            mapping: report.versions.mapping,
+            sanitizer: report.versions.sanitizer,
+            parser: report.versions.parser,
+          })
+          expect(Object.keys(report.versions).sort()).toStrictEqual(['adapter', 'mapping', 'parser', 'sanitizer'])
+          expect(typeof report.adapterTiming.parseMs).toBe('number')
+          expect(typeof report.adapterTiming.sanitizeMs).toBe('number')
+        } finally {
+          await cleanupSiteRows(db, siteId)
+        }
+      } finally {
+        await rm(w.work, { recursive: true, force: true })
+      }
+    })
+  })
+
+  it('二导幂等（换 key）：全部写入计数归零、数据库行数不变、job 仍 succeeded_with_issues（降级汇总 issue 必在）', async () => {
+    await withTestDb(async (db) => {
+      const w = await makeWork('sf-full-idem')
+      try {
+        const { key, siteId, wxrId } = await registerFull(db, w)
+        try {
+          const base = { siteKey: key, snapshotId: wxrId, reportRoot: w.reportRoot } as const
+          const first = await createImportJob(db, { ...base })
+          expect(first.status).toBe('succeeded_with_issues')
+
+          const again = await createImportJob(db, { ...base, idempotencyKey: `sffull-rerun-${rand()}` })
+          expect(again.reused).toBe(false)
+          expect(again.status).toBe('succeeded_with_issues')
+
+          // 写入计数全零语义：按事件计的键（termsProcessed/mediaDanglingRefs）如实再现
+          const report = await readReport(w, again.jobId)
+          expect(report.counts).toStrictEqual({
+            contentEntries: 0,
+            revisionsCreated: 0,
+            sourceRecords: 0,
+            skipped: 3,
+            termsProcessed: 2,
+            autoCreatedTerms: 0,
+            contentTermsCreated: 0,
+            mediaRegistered: 0,
+            mediaFromSidecar: 0,
+            mediaMissing: 0,
+            mediaQuarantined: 0,
+            mediaDanglingRefs: 2,
+            mediaAmbiguous: 0,
+            routesCreated: 0,
+            routesUpdated: 0,
+            routesRedirects: 0,
+            routesCollisions: 0,
+            routesExcludedQueries: 0,
+          })
+
+          // 数据库行数不变（幂等：零新行零修订零更新）
+          expect(await countOf(db, `select count(*)::int as n from content_entries where site_id = '${siteId}'`)).toBe(3)
+          expect(await countOf(db, `select count(*)::int as n from content_revisions where site_id = '${siteId}'`)).toBe(3)
+          expect(await countOf(db, `select count(*)::int as n from source_records where site_id = '${siteId}'`)).toBe(5)
+          expect(await countOf(db, `select count(*)::int as n from terms where site_id = '${siteId}'`)).toBe(2)
+          expect(await countOf(db, `select count(*)::int as n from content_terms where site_id = '${siteId}'`)).toBe(1)
+          expect(await countOf(db, `select count(*)::int as n from media_assets where site_id = '${siteId}'`)).toBe(2)
+          expect(await countOf(db, `select count(*)::int as n from media_usages where site_id = '${siteId}'`)).toBe(3)
+          expect(await countOf(db, `select count(*)::int as n from route_registry where site_id = '${siteId}'`)).toBe(12)
+
+          // 降级汇总 issue 必在（Plan 3 钉死：降级是处置事实，重跑如实再现）+ 悬空 2 条
+          const downgrade = report.issues.filter((i: string) => i.includes('attachment 页降级'))
+          expect(downgrade).toHaveLength(1)
+          expect(downgrade[0]).toContain('2 条')
+          expect(report.issues.filter((i: string) => i.includes('悬空'))).toHaveLength(2)
+          expect(report.issues).toHaveLength(3)
+        } finally {
+          await cleanupSiteRows(db, siteId)
+        }
+      } finally {
+        await rm(w.work, { recursive: true, force: true })
+      }
+    })
+  })
+
+  it('无 media 快照（只 wxr）：media 域全 missing 报告不失败（空 manifest 语义）+ 主列表路径参数入报告', async () => {
+    await withTestDb(async (db) => {
+      const w = await makeWork('sf-full-nomedia')
+      try {
+        const { key, siteId, wxrId } = await registerWxrOnly(db, w)
+        try {
+          const result = await createImportJob(db, {
+            siteKey: key,
+            snapshotId: wxrId,
+            reportRoot: w.reportRoot,
+            productsPath: '/shop/',
+          })
+          expect(result.status).toBe('succeeded_with_issues')
+
+          // 空清单语义：2 attachment 全 missing（不插行）、5 引用全悬空、路由 8 行
+          // （3 站点页 + 3 内容 + 2 分类——301/asset 无资产可指）
+          expect(await countOf(db, `select count(*)::int as n from media_assets where site_id = '${siteId}'`)).toBe(0)
+          expect(await countOf(db, `select count(*)::int as n from route_registry where site_id = '${siteId}'`)).toBe(8)
+          expect(
+            await countOf(db, `select count(*)::int as n from route_registry where site_id = '${siteId}' and route_path = '/shop/'`),
+          ).toBe(1)
+          expect(
+            await countOf(db, `select count(*)::int as n from route_registry where site_id = '${siteId}' and route_path = '/products/'`),
+          ).toBe(0)
+
+          const report = await readReport(w, result.jobId)
+          expect(report.counts.mediaMissing).toBe(2)
+          expect(report.counts.mediaRegistered).toBe(0)
+          expect(report.counts.mediaDanglingRefs).toBe(5)
+          expect(report.counts.routesCreated).toBe(8)
+          expect(report.paths).toStrictEqual({ productsPath: '/shop/', newsPath: '/news/' })
+          // attachment 301 目标缺席逐条报告（§7.6）+ missing/悬空——全部在 issues
+          expect(report.issues.filter((i: string) => i.includes('301 目标缺席'))).toHaveLength(2)
+          expect(report.issues.filter((i: string) => i.includes('missing'))).toHaveLength(2)
+          expect(report.issues.filter((i: string) => i.includes('悬空'))).toHaveLength(5)
+        } finally {
+          await cleanupSiteRows(db, siteId)
+        }
+      } finally {
+        await rm(w.work, { recursive: true, force: true })
+      }
+    })
+  })
+
+  it('NULL-entry 占位行收编（Task 3 审查移交①）：六列命中 content_entry_id NULL 行 → UPDATE 挂 entry、不撞 23505、sourceRecords 不虚增', async () => {
+    await withTestDb(async (db) => {
+      const w = await makeWork('sf-full-adopt')
+      try {
+        const { key, siteId, wxrId } = await registerWxrOnly(db, w)
+        try {
+          // 外源占位行（seed 模拟：六列身份 = wp:post/101 本快照、entry NULL）
+          const seedJob = (await db.execute(`insert into import_jobs (id, site_id, source_snapshot_id, status,
+            mapping_version, idempotency_key, attempt) values (gen_random_uuid(), '${siteId}', '${wxrId}',
+            'succeeded', 'seed', 'seed-${rand()}', 1) returning id`)) as any
+          const connRow = (await db.execute(`select source_connection_id from source_snapshots where id = '${wxrId}'`)) as any
+          const seeded = (await db.execute(`insert into source_records (id, site_id, source_connection_id,
+            source_snapshot_id, import_job_id, source_type, source_id, locale, raw_location, parse_status,
+            content_entry_id) values (gen_random_uuid(), '${siteId}', '${connRow[0].source_connection_id}',
+            '${wxrId}', '${seedJob[0].id}', 'wp:post', '101', 'en', 'wxr://item/101', 'imported', NULL)
+            returning id`)) as any
+
+          const result = await createImportJob(db, { siteKey: key, snapshotId: wxrId, reportRoot: w.reportRoot })
+          expect(result.status).toBe('succeeded_with_issues')
+
+          // 占位行被收编：挂 entry + checksum 前移（非新建行——sourceRecords=4 不是 5）
+          const adopted = (await db.execute(`select content_entry_id, normalized_checksum from source_records
+            where id = '${seeded[0].id}'`)) as any
+          expect(adopted[0].content_entry_id).not.toBeNull()
+          expect(adopted[0].normalized_checksum).not.toBeNull()
+          const el = (await db.execute(`select current_revision_id, published_revision_id from entry_locales
+            where site_id = '${siteId}' and content_entry_id = '${adopted[0].content_entry_id}'`)) as any
+          expect(el[0].current_revision_id).not.toBeNull()
+          expect(el[0].published_revision_id).toBeNull()
+
+          const report = await readReport(w, result.jobId)
+          expect(report.counts.contentEntries).toBe(3)
+          expect(report.counts.revisionsCreated).toBe(3)
+          expect(report.counts.sourceRecords).toBe(4) // 101 收编非新建（2 attachment + 111/121）
+          expect(report.counts.skipped).toBe(0)
+          // 收编行照常参与 term 关联与路由（101 的 entry 锚点就位）
+          expect(report.counts.contentTermsCreated).toBe(1)
+          expect(report.counts.routesCreated).toBe(8)
+        } finally {
+          await cleanupSiteRows(db, siteId)
+        }
+      } finally {
+        await rm(w.work, { recursive: true, force: true })
+      }
+    })
+  })
+
+  it('stale 回收（义务④）：同 site 超 24h 的 queued/running 置 stale；新近 job 不受影响；新 job 正常推进', async () => {
+    await withTestDb(async (db) => {
+      const w = await makeWork('sf-full-stale')
+      try {
+        const { key, siteId, wxrId } = await registerWxrOnly(db, w)
+        try {
+          const oldH = 25
+          const freshH = 1
+          const seed = async (status: string, hoursAgo: number, idem: string): Promise<string> => {
+            const rows = (await db.execute(`insert into import_jobs (id, site_id, source_snapshot_id, status,
+              mapping_version, idempotency_key, attempt, started_at, created_at) values (gen_random_uuid(),
+              '${siteId}', '${wxrId}', '${status}', 'seed', '${idem}', 1,
+              now() - interval '${hoursAgo} hours', now() - interval '${hoursAgo} hours') returning id`)) as any
+            return rows[0].id
+          }
+          const oldRunning = await seed('running', oldH, `stale-run-${rand()}`)
+          const oldQueued = await seed('queued', oldH + 2, `stale-q-${rand()}`)
+          const freshRunning = await seed('running', freshH, `stale-fresh-${rand()}`)
+
+          const result = await createImportJob(db, {
+            siteKey: key,
+            snapshotId: wxrId,
+            reportRoot: w.reportRoot,
+            pipeline: async () => ({ counts: {}, issues: [] }),
+          })
+          expect(result.status).toBe('succeeded')
+
+          const st = (rows: string[]) =>
+            (db.execute(`select status, finished_at from import_jobs where id in (${rows.map((r) => `'${r}'`).join(',')})`) as any) as Promise<any>
+          const oldRows = await st([oldRunning, oldQueued])
+          for (const r of oldRows) {
+            expect(r.status).toBe('stale')
+            expect(r.finished_at).not.toBeNull()
+          }
+          const freshRows = await st([freshRunning])
+          expect(freshRows[0].status).toBe('running') // 未超时不动（心跳列决策留 ADR-017）
+          expect(freshRows[0].finished_at).toBeNull()
+        } finally {
+          await cleanupSiteRows(db, siteId)
+        }
+      } finally {
+        await rm(w.work, { recursive: true, force: true })
+      }
+    })
+  })
+
+  it('REPORT_WRITE_FAILED（Task 1 遗留）：报告目录不可建 → failed 终态、error_code 落列、report_path 不落', async () => {
+    await withTestDb(async (db) => {
+      const w = await makeWork('sf-full-rwfail')
+      try {
+        const { key, siteId, wxrId } = await registerWxrOnly(db, w)
+        try {
+          // 用普通文件占住报告根：mkdir <file>/<jobId>/reports 必败（ENOTDIR/EEXIST）
+          const blocker = join(w.work, 'blocker')
+          await writeFile(blocker, 'not a directory')
+          const result = await createImportJob(db, {
+            siteKey: key,
+            snapshotId: wxrId,
+            reportRoot: blocker,
+            pipeline: async () => ({ counts: { ok: 1 }, issues: [] }),
+          })
+          expect(result.status).toBe('failed')
+          expect(result.reportPath).toBeNull()
+          const job = (await db.execute(`select status, error_code, error_summary, report_path, finished_at
+            from import_jobs where id = '${result.jobId}'`)) as any
+          expect(job[0].status).toBe('failed')
+          expect(job[0].error_code).toBe('REPORT_WRITE_FAILED')
+          expect(String(job[0].error_summary)).toMatch(/^\[report\] /)
+          expect(job[0].report_path).toBeNull() // 文件不存在不悬挂路径
+          expect(job[0].finished_at).not.toBeNull()
+        } finally {
+          await cleanupSiteRows(db, siteId)
+        }
+      } finally {
+        await rm(w.work, { recursive: true, force: true })
+      }
+    })
+  })
+})
diff --git a/tests/integration/import-job.test.ts b/tests/integration/import-job.test.ts
index b11fd7b..fe9eb13 100644
--- a/tests/integration/import-job.test.ts
+++ b/tests/integration/import-job.test.ts
@@ -11,7 +11,7 @@
 //    显式 id 不存在或属他站 → SNAPSHOT_NOT_FOUND；站点不存在 → SITE_NOT_FOUND；均零 job 行；
 // 5. failed 路径：管道抛错 → running→failed（error_code/error_summary/finished_at，
 //    '[pipeline]' 前缀），重提交同 key 复用失败 job 不重跑；管道抛 ImportError 保留原 code；
-// 6. Outcome 契约：ctx.report 累加 ⊕ 返回值（counts 同名键返回值覆盖）→ succeeded_with_issues；
+// 6. Outcome 契约：ctx.report 累加 ⊕ 返回值（counts 同名键**求和**——Task 6 编排聚合）→ succeeded_with_issues；
 // 7. 复用口径忠实化：reuse 返回只含 job 行锚定字段（快照/版本字段缺席）；
 // 8. 23505 收敛确定性探针：未提交孪生 job 持有幂等键（真实 PG 锁语义，零代理）。
 // 运行于 globalSetup 临时库（ADR-015）；storage/报告根用 os.tmpdir 临时目录，finally 清理。
@@ -434,7 +434,7 @@ describe('import_jobs 生命周期（Task 1 骨架）', () => {
     })
   })
 
-  it('Outcome 契约：ctx.report 累加 ⊕ 返回值（返回值 counts 覆盖同名键），issues 非空 → succeeded_with_issues，计数与问题入报告', async () => {
+  it('Outcome 契约：ctx.report 累加 ⊕ 返回值按**键求和**（Task 6 编排聚合），issues 非空 → succeeded_with_issues，计数与问题入报告', async () => {
     await withTestDb(async (db) => {
       const ctx = await makeCtx()
       try {
@@ -449,7 +449,7 @@ describe('import_jobs 生命周期（Task 1 骨架）', () => {
             pipeline: async (jobCtx) => {
               // 写入器式累加（Task 6 写入器的使用形态）
               jobCtx.report.counts['terms'] = 2
-              jobCtx.report.counts['contentEntries'] = 999 // 被返回值同名键覆盖
+              jobCtx.report.counts['contentEntries'] = 100 // 与返回值同名键**求和**（Task 6 钉死：防覆盖安全网）
               jobCtx.report.issues.push('自建缺失 term：1（category/uncategorized）')
               return {
                 counts: { contentEntries: 3, routes: 8 },
@@ -468,8 +468,8 @@ describe('import_jobs 生命周期（Task 1 骨架）', () => {
             await readFile(join(ctx.importReportRoot, result.jobId, 'reports', 'import.json'), 'utf8'),
           ) as { status: string; counts: Record<string, number>; issues: string[] }
           expect(report.status).toBe('succeeded_with_issues')
-          // 汇合规则：counts 按键合并（返回值覆盖累加器同名键）、issues 累加器在前返回值在后
-          expect(report.counts).toStrictEqual({ terms: 2, contentEntries: 3, routes: 8 })
+          // 汇合规则：counts 按键**求和**（Task 6 钉死——CLOBBER 废弃）、issues 累加器在前返回值在后
+          expect(report.counts).toStrictEqual({ terms: 2, contentEntries: 103, routes: 8 })
           expect(report.issues).toStrictEqual([
             '自建缺失 term：1（category/uncategorized）',
             '悬空 attachment 引用：2 条（mini 203/204）',