feat: add notification-threshold rollback apply guardrails

Author: Jacobinwwey

docs/brainstorms/2026-04-16-mainline-ci-stabilization-and-m7-direction-requirements.md

@@ -792,7 +792,7 @@ Suggested verification slice:
 - Best next increment is not broader governance expansion.
   Best next increment is explicit rollback execution guardrails on top of the preview surface.
 
-### M7.24 (Next): Notification-Threshold Rollback Apply Guardrails (Lane Ops Bridge)
+### M7.24 (Now): Notification-Threshold Rollback Apply Guardrails (Lane Ops Bridge)
 
 Deliverables:
 
@@ -808,6 +808,62 @@ Suggested verification slice:
 - `npm run docs:diataxis:check`
 - `npm run docs:site:build`
 
+#### M7.24 Progress Note (2026-04-16)
+
+- [Done] expanded `src/server.ts` with explicit rollback-apply route:
+  - `POST /api/knowledge/operator/agent-workspace-diagnostics/triage/remediation/escalation/notification-thresholds/apply`.
+- [Done] added deterministic preview-to-apply guardrails:
+  - `buildAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreviewFingerprint(...)`,
+  - `applyAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreview(...)`,
+  - `normalizeAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreviewMode(...)`.
+- [Done] preview payloads now carry explicit execution fingerprints:
+  - `preview.resetToDefault.previewFingerprint`,
+  - `preview.rollbackToPrevious.previewFingerprint`.
+- [Done] apply semantics are now compare-and-swap style instead of blind mutation:
+  - reset/apply requests must send the reviewed preview fingerprint,
+  - rollback-to-previous requests must also send the reviewed `auditId`,
+  - stale preview fingerprints, unavailable rollback previews, and mismatched audit ids are blocked with deterministic `409` responses,
+  - no-op applies against already-active targets are blocked instead of silently writing redundant audit entries.
+- [Done] apply response now returns bounded execution evidence:
+  - `appliedMode`,
+  - `previewFingerprint`,
+  - `matchedPreview`,
+  - `appliedAuditEntry`.
+- [Done] expanded evidence coverage:
+  - `src/server.migration.test.ts` now validates audit-id mismatch blocking, stale fingerprint blocking, rollback apply success, reset apply success, and audit-trail ordering after both mutations.
+  - `src/knowledge.api.contract.test.ts`, `src/agent_workspace.verification.contract.test.ts`, and `scripts/verify-agent-workspace-runtime.js` now fail fast on apply-route and fingerprint/apply helper drift.
+- [Done] verification evidence:
+  - `npm test -- src/server.migration.test.ts --runInBand --testNamePattern "notification threshold rollback apply guardrails stay deterministic"`
+  - `npm run test:agent-workspace:contracts`
+  - `npm run verify:agent-workspace:runtime`
+  - `npm run docs:diataxis:check`
+  - `npm run docs:site:build`
+
+### Post-M7.24 Architecture Judgment
+
+- L5 governance now supports bounded preview-to-apply execution for notification threshold rollback and reset.
+- Current weak point shifts from mutation safety to post-apply reconciliation:
+  - operators can now preview and apply deterministic targets,
+  - operators still lack a dedicated readback surface that says whether the latest applied preview remains current or has already been superseded by later threshold changes.
+- Best next increment is not broader remediation scope.
+  Best next increment is a narrow apply-outcome reconciliation surface on top of existing audit artifacts.
+
+### M7.25 (Next): Notification-Threshold Apply Outcome Reconciliation (Lane Ops Bridge)
+
+Deliverables:
+
+- add read-only reconciliation surface for the latest notification-threshold apply outcome.
+- compare latest applied preview fingerprint, latest audit entry, and current active thresholds so operators can see whether the reviewed/applyed state is still current or has drifted.
+- keep scope limited to notification-threshold governance; do not reopen UI, broader remediation, or unrelated memory/retrieval lanes.
+
+Suggested verification slice:
+
+- `npm test -- src/server.migration.test.ts --runInBand --testNamePattern "notification threshold apply outcome reconciliation"`
+- `npm run test:agent-workspace:contracts`
+- `npm run verify:agent-workspace:runtime`
+- `npm run docs:diataxis:check`
+- `npm run docs:site:build`
+
 ## Success Criteria
 
 - CI failure mode that previously blocked the three agent-workspace suites is eliminated on mainline.
@@ -817,4 +873,4 @@ Suggested verification slice:
 
 ## Next Step
 
-Proceed to `/prompts:ce-plan` using this document as the source for `M7.24` decomposition (notification-threshold rollback apply guardrails), while preserving M7 lane boundary constraints.
+Proceed to `/prompts:ce-plan` using this document as the source for `M7.25` decomposition (notification-threshold apply outcome reconciliation), while preserving M7 lane boundary constraints.

docs/diataxis/en/explanation/development-progress-dashboard.md

@@ -599,29 +599,32 @@ Execution anchor:
   - `npm run test:agent-workspace:contracts`
   - `npm run verify:agent-workspace:runtime`
 
-## Latest Mainline Increment (2026-04-16 M7.23 Notification-Threshold Rollback Preview and Drift-Diff Governance Lane)
-
-- Expanded `src/server.ts` with preview route:
-  - `GET /api/knowledge/operator/agent-workspace-diagnostics/triage/remediation/escalation/notification-thresholds/preview`.
-- Added deterministic rollback-preview helper stack:
-  - `buildAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreview(...)`,
-  - `areAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPoliciesEquivalent(...)`.
-- Hardened operator visibility without widening scope:
-  - preview payload now exposes `activeThresholds`, `defaultThresholds`, `driftDiff.activeVsDefault`, `driftDiff.activeVsLatestAuditPrevious`, `preview.resetToDefault`, and `preview.rollbackToPrevious`.
-- Added latest-audit-linked rollback semantics:
-  - rollback-to-previous preview is only marked available when current active thresholds still match the latest audited `nextThresholds`,
-  - drift remains read-only when preview and active state no longer align.
+## Latest Mainline Increment (2026-04-16 M7.24 Notification-Threshold Rollback Apply Guardrails Lane)
+
+- Expanded `src/server.ts` with explicit apply route:
+  - `POST /api/knowledge/operator/agent-workspace-diagnostics/triage/remediation/escalation/notification-thresholds/apply`.
+- Added deterministic preview-to-apply helper stack:
+  - `buildAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreviewFingerprint(...)`,
+  - `applyAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreview(...)`,
+  - `normalizeAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreviewMode(...)`.
+- Hardened mutation safety without widening scope:
+  - preview payloads now expose `preview.resetToDefault.previewFingerprint` and `preview.rollbackToPrevious.previewFingerprint`,
+  - apply requests must match the reviewed fingerprint,
+  - rollback-to-previous apply requests must also match the reviewed `auditId`,
+  - stale fingerprints, unavailable rollback previews, mismatched audit ids, and no-op targets are rejected deterministically.
+- Added bounded execution receipts:
+  - apply responses now return `appliedMode`, `previewFingerprint`, `matchedPreview`, and `appliedAuditEntry`.
 - Expanded executable evidence:
-  - `src/server.migration.test.ts` now validates preview-route payloads, default-vs-active drift deltas, latest-audit rollback deltas, and reset-preview deltas.
+  - `src/server.migration.test.ts` now validates audit-id mismatch blocking, stale fingerprint blocking, rollback apply success, reset apply success, and persisted audit ordering.
 - Hardened runtime verification gate:
-  - `src/knowledge.api.contract.test.ts`, `src/agent_workspace.verification.contract.test.ts`, and `scripts/verify-agent-workspace-runtime.js` now fail fast on notification-threshold preview route and helper drift.
+  - `src/knowledge.api.contract.test.ts`, `src/agent_workspace.verification.contract.test.ts`, and `scripts/verify-agent-workspace-runtime.js` now fail fast on notification-threshold apply route and fingerprint/apply helper drift.
 - Verification evidence:
-  - `npm test -- src/server.migration.test.ts --runInBand --testNamePattern \"escalation notification threshold rollback preview and drift-diff governance stay deterministic\"`
+  - `npm test -- src/server.migration.test.ts --runInBand --testNamePattern \"notification threshold rollback apply guardrails stay deterministic\"`
   - `npm run test:agent-workspace:contracts`
   - `npm run verify:agent-workspace:runtime`
 - Next direction judgment:
-  - current bottleneck shifts from visibility to controlled rollback execution,
-  - M7.24 should stay bounded to rollback-apply guardrails with preview fingerprinting, not broader governance expansion.
+  - current bottleneck shifts from guarded execution to post-apply reconciliation,
+  - M7.25 should stay bounded to apply-outcome reconciliation, not broader remediation or UI expansion.
 
 ## Mainline vs Working-Branch Snapshot (2026-04-14)
 
@@ -671,7 +674,7 @@ This dashboard aligns against the following requirement chain:
 | L2 Retrieval | explainable hybrid/vector retrieval + governance | Expanded in branch-oriented plans | Mainline file-backed baseline only (`src/learning/store.ts`) | Re-enter lane after concrete module evidence lands on mainline |
 | L3 Learning | mastery diagnostics + path/session loop | Expanded in branch | Partially integrated | Contract and integration parity |
 | L4 Interaction | agent conversation + focus/path pane runtime | Implemented in branch | M1-M4 baseline integrated on mainline | Expand capability surface via typed contract only |
-| L5 Governance | runbook, diagnostics, replay/autonomy controls | Expanded in branch | Operator diagnostics persistence/triage/history/threshold governance + runbook automation/audit + adaptive simulation/remediation + remediation backtest/approval-gate + approval-policy hardening/regression-alarms + approval-policy drift/escalation + escalation acknowledgement lifecycle/audit + escalation SLA/reminder baseline + notification digest/suppression baseline + delivery-log observability + stale-cleanup health auditing + anomaly/retention governance + notification SLO governance + notification-threshold override/audit governance + rollback preview/drift-diff governance integrated | M7.24: notification-threshold rollback apply guardrails |
+| L5 Governance | runbook, diagnostics, replay/autonomy controls | Expanded in branch | Operator diagnostics persistence/triage/history/threshold governance + runbook automation/audit + adaptive simulation/remediation + remediation backtest/approval-gate + approval-policy hardening/regression-alarms + approval-policy drift/escalation + escalation acknowledgement lifecycle/audit + escalation SLA/reminder baseline + notification digest/suppression baseline + delivery-log observability + stale-cleanup health auditing + anomaly/retention governance + notification SLO governance + notification-threshold override/audit governance + rollback preview/drift-diff governance + rollback apply guardrails integrated | M7.25: notification-threshold apply outcome reconciliation |
 
 ## Verification Baseline
 

docs/diataxis/zh/explanation/development-progress-dashboard.md

@@ -601,29 +601,32 @@
   - `npm run test:agent-workspace:contracts`
   - `npm run verify:agent-workspace:runtime`
 
-## 主线最新增量（2026-04-16 M7.23 通知阈值回滚预览与 drift-diff 治理链路）
-
-- 已在 `src/server.ts` 增加预览路由：
-  - `GET /api/knowledge/operator/agent-workspace-diagnostics/triage/remediation/escalation/notification-thresholds/preview`。
-- 已增加确定性回滚预览 helper 栈：
-  - `buildAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreview(...)`，
-  - `areAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPoliciesEquivalent(...)`。
-- 已在不扩 scope 前提下补足 operator 可见性：
-  - 预览载荷现在输出 `activeThresholds`、`defaultThresholds`、`driftDiff.activeVsDefault`、`driftDiff.activeVsLatestAuditPrevious`、`preview.resetToDefault` 与 `preview.rollbackToPrevious`。
-- 已增加 latest-audit-linked rollback 语义：
-  - 仅当当前 active thresholds 仍与最新审计记录的 `nextThresholds` 对齐时，rollback-to-previous preview 才标记为可用，
-  - 否则只暴露只读 drift 信息，不做隐式回滚。
+## 主线最新增量（2026-04-16 M7.24 通知阈值回滚 apply guardrails 链路）
+
+- 已在 `src/server.ts` 增加显式 apply 路由：
+  - `POST /api/knowledge/operator/agent-workspace-diagnostics/triage/remediation/escalation/notification-thresholds/apply`。
+- 已增加确定性 preview-to-apply helper 栈：
+  - `buildAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreviewFingerprint(...)`，
+  - `applyAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreview(...)`，
+  - `normalizeAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreviewMode(...)`。
+- 已在不扩 scope 前提下加固 mutation 安全：
+  - 预览载荷现在输出 `preview.resetToDefault.previewFingerprint` 与 `preview.rollbackToPrevious.previewFingerprint`，
+  - apply 请求必须匹配已审阅的 preview fingerprint，
+  - rollback-to-previous apply 请求还必须匹配已审阅的 `auditId`，
+  - stale fingerprint、不可用 rollback preview、auditId 不匹配与 no-op target 都会被确定性拒绝。
+- 已增加有界执行回执：
+  - apply 响应现在返回 `appliedMode`、`previewFingerprint`、`matchedPreview` 与 `appliedAuditEntry`。
 - 已补可执行证据：
-  - `src/server.migration.test.ts` 新增 preview 路由载荷、default-vs-active drift delta、latest-audit rollback delta 与 reset preview delta 断言。
+  - `src/server.migration.test.ts` 现在覆盖 auditId mismatch 阻断、stale fingerprint 阻断、rollback apply 成功、reset apply 成功与持久化 audit 顺序断言。
 - 已加固 runtime 门禁：
-  - `src/knowledge.api.contract.test.ts`、`src/agent_workspace.verification.contract.test.ts` 与 `scripts/verify-agent-workspace-runtime.js` 新增 notification-threshold preview 路由与 helper 的 fail-fast 断言。
+  - `src/knowledge.api.contract.test.ts`、`src/agent_workspace.verification.contract.test.ts` 与 `scripts/verify-agent-workspace-runtime.js` 现在对 notification-threshold apply 路由与 fingerprint/apply helper 做 fail-fast 断言。
 - 验证证据：
-  - `npm test -- src/server.migration.test.ts --runInBand --testNamePattern \"escalation notification threshold rollback preview and drift-diff governance stay deterministic\"`
+  - `npm test -- src/server.migration.test.ts --runInBand --testNamePattern \"notification threshold rollback apply guardrails stay deterministic\"`
   - `npm run test:agent-workspace:contracts`
   - `npm run verify:agent-workspace:runtime`
 - 后续方向判断：
-  - 当前瓶颈从可见性转向受控执行，
-  - M7.24 应继续收敛在 rollback apply guardrails 与 preview fingerprinting，不要扩成更大的治理子系统。
+  - 当前瓶颈从受控执行转向 post-apply reconciliation，
+  - M7.25 应继续收敛在 apply 结果对账，不要扩成更大的 remediation 或 UI 子系统。
 
 ## 主线 vs 工作分支快照（2026-04-14）
 
@@ -673,7 +676,7 @@
 | L2 检索层 | 可解释混合/向量检索 + 治理 | 分支规划增强中 | 主线当前为 file-backed 基线（`src/learning/store.ts`） | 待主线出现对应模块证据后再收敛 |
 | L3 学习层 | 掌握诊断 + 路径/会话闭环 | 分支增强中 | 主线部分集成 | 契约与集成一致性 |
 | L4 交互层 | agent 对话 + focus/path pane 运行时 | 分支已实现 | 主线 M1-M4 已落入基线 | 继续通过 typed contract 扩展动作面 |
-| L5 治理层 | runbook/诊断/回放与自动化 | 分支增强中 | 主线已集成运维诊断持久化/分级/趋势历史/阈值治理 + runbook 自动化/阈值审计 + 自适应模拟/自动修复 + 回测/批准门禁 + 批准策略硬化/回归告警 + 批准策略漂移/升级 + 升级确认生命周期/审计 + 升级 SLA/提醒基线 + 通知摘要/抑制基线 + 交付日志可观测性 + 陈旧通知健康审计 + 异常/retention 治理 + 通知 SLO 治理 + 通知阈值覆盖/审计治理 + 回滚预览/drift-diff 治理 | M7.24：通知阈值回滚 apply guardrails |
+| L5 治理层 | runbook/诊断/回放与自动化 | 分支增强中 | 主线已集成运维诊断持久化/分级/趋势历史/阈值治理 + runbook 自动化/阈值审计 + 自适应模拟/自动修复 + 回测/批准门禁 + 批准策略硬化/回归告警 + 批准策略漂移/升级 + 升级确认生命周期/审计 + 升级 SLA/提醒基线 + 通知摘要/抑制基线 + 交付日志可观测性 + 陈旧通知健康审计 + 异常/retention 治理 + 通知 SLO 治理 + 通知阈值覆盖/审计治理 + 回滚预览/drift-diff 治理 + 回滚 apply guardrails | M7.25：通知阈值 apply 结果对账 |
 
 ## 验证基线
 

scripts/verify-agent-workspace-runtime.js

@@ -164,6 +164,10 @@ function verifyAgentWorkspaceRuntime(repoRoot = path.resolve(__dirname, '..')) {
     serverSource.includes('/api/knowledge/operator/agent-workspace-diagnostics/triage/remediation/escalation/notification-thresholds/preview'),
     'Missing diagnostics remediation escalation notification-threshold preview route in src/server.ts'
   );
+  assert(
+    serverSource.includes('/api/knowledge/operator/agent-workspace-diagnostics/triage/remediation/escalation/notification-thresholds/apply'),
+    'Missing diagnostics remediation escalation notification-threshold apply route in src/server.ts'
+  );
   assert(
     serverSource.includes('/api/knowledge/operator/agent-workspace-diagnostics/triage/remediation/escalation/notification-thresholds/audit'),
     'Missing diagnostics remediation escalation notification-threshold audit route in src/server.ts'
@@ -348,6 +352,14 @@ function verifyAgentWorkspaceRuntime(repoRoot = path.resolve(__dirname, '..')) {
     serverSource.includes('buildAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreview'),
     'Missing remediation escalation notification threshold preview helper in src/server.ts'
   );
+  assert(
+    serverSource.includes('buildAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreviewFingerprint'),
+    'Missing remediation escalation notification threshold preview fingerprint helper in src/server.ts'
+  );
+  assert(
+    serverSource.includes('applyAgentWorkspaceDiagnosticsRemediationEscalationNotificationThresholdPreview'),
+    'Missing remediation escalation notification threshold apply helper in src/server.ts'
+  );
   assert(
     serverSource.includes('buildAgentWorkspaceDiagnosticsRemediationEscalationNotificationAnomalyReport'),
     'Missing remediation escalation notification anomaly report helper in src/server.ts'