bmad-speckit-sdd-flow 1.0.2 → 1.1.0

package/README.md

@@ -7,7 +7,7 @@ English | [简体中文](README.zh-CN.md)
 </p>
 
 <h3 align="center">
-  Governed Spec-Driven AI Delivery for Cursor and Claude Code
+  Governed Spec-Driven AI Delivery for Cursor, Claude Code, and Codex
 </h3>
 
 <p align="center">
@@ -46,10 +46,11 @@ Traditional AI tooling often stops at prompt orchestration. BMAD-Speckit-SDD-Flo
 ## Runtime Governance At A Glance
 
 - **Four-Signal readiness gate** runs before implementation entry and keeps readiness scoring separate from implementation scoring.
-- **Runtime gates loop** keeps workflow progression aligned with stage truth instead of letting hosts continue optimistically.
-- **Rerun gates** carry remediation and re-entry through the same governed path after fixes.
-- **Packet execution closure** preserves evidence for every governed execution outcome across Cursor and Claude Code hosts.
-- **Post-audit close-out** feeds scoring, dashboard, coach, and SFT extraction from the same runtime close-out path.
+- **`main-agent-orchestration inspect` is the first read**: the main agent reads the repo-native authoritative surface before choosing the next global branch.
+- **`dispatch-plan` is conditional**: only materialize a plan when the surface says the next branch should dispatch a packet.
+- **Child agents only execute a bounded packet**: they return packet results, but they do not choose the next global branch.
+- **`runAuditorHost` is post-audit close-out only**: it closes approved audits into scoring, dashboard, coach, and SFT outputs, then the main agent re-reads `inspect`.
+- **Legacy worker/manual close-out playbooks are historical evidence**: keep them for auditability, but they are not the active accepted runtime path.
 
 ## Dashboard And MCP
 
@@ -78,6 +79,7 @@ This is the verified off-repo path for the published package contract.
 npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit version
 npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit-init . --agent claude-code --full --no-package-json
 npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit-init . --agent cursor --full --no-package-json
+npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit-init . --agent codex --full --no-package-json
 npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit check
 npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit dashboard-status
 ```
@@ -164,16 +166,16 @@ This only removes installer-managed entries. It does not delete `.cursor`, `.cla
 
 ### Core Components
 
-| Component                   | Purpose                                                                                             |
-| :-------------------------- | :-------------------------------------------------------------------------------------------------- |
-| **`_bmad/`**                | Canonical source of truth for workflow modules, hooks, prompts, routing, and host-facing assets.    |
-| **`packages/scoring/`**     | Scoring engine, readiness drift evaluation, dashboard projection, coach inputs, and SFT extraction. |
-| **`dashboard`**             | Default runtime observability surface: live dashboard, runtime snapshots, and scoring projections.  |
-| **`runtime-mcp`**           | Optional MCP surface for agent tools over runtime data; enabled explicitly with `--with-mcp`.       |
-| **`speckit-workflow`**      | Specify -> Plan -> GAPS -> Tasks -> TDD with mandatory audit loops.                                 |
-| **`bmad-story-assistant`**  | Story lifecycle path: Create Story -> Party Mode -> Dev Story -> Implement.                         |
-| **`bmad-bug-assistant`**    | Bug lifecycle path: RCA -> Party Mode -> BUGFIX -> Implement.                                       |
-| **`bmad-standalone-tasks`** | Standalone execution path for TASKS or BUGFIX documents through governed subagent delivery.         |
+| Component                   | Purpose                                                                                                                                    |
+| :-------------------------- | :----------------------------------------------------------------------------------------------------------------------------------------- |
+| **`_bmad/`**                | Canonical source of truth for workflow modules, hooks, prompts, routing, and host-facing assets.                                           |
+| **`packages/scoring/`**     | Scoring engine, readiness drift evaluation, dashboard projection, coach inputs, and SFT extraction.                                        |
+| **`dashboard`**             | Default runtime observability surface: live dashboard, runtime snapshots, and scoring projections.                                         |
+| **`runtime-mcp`**           | Optional MCP surface for agent tools over runtime data; enabled explicitly with `--with-mcp`.                                              |
+| **`speckit-workflow`**      | Specify -> Plan -> GAPS -> Tasks -> TDD with mandatory audit loops.                                                                        |
+| **`bmad-story-assistant`**  | Story lifecycle entry: main agent reads `inspect`, dispatches bounded packets when needed, and closes post-audit through `runAuditorHost`. |
+| **`bmad-bug-assistant`**    | Bug lifecycle path: RCA -> Party Mode -> BUGFIX -> Implement, while the main Agent still owns the global `inspect -> dispatch-plan -> closeout` chain. |
+| **`bmad-standalone-tasks`** | TASKS/BUGFIX execution still follows the main-agent path: `inspect` first, `dispatch-plan` only when needed, then bounded subagent work.             |
 
 <details>
 <summary><b>View Folder Structure</b></summary>
@@ -197,6 +199,7 @@ BMAD-Speckit-SDD-Flow/
 Key entry points:
 
 - [Getting Started](docs/tutorials/getting-started.md)
+- [Main-Agent Orchestration Reference](docs/reference/main-agent-orchestration.md)
 - [Consumer Installation Guide](docs/how-to/consumer-installation.md)
 - [Runtime Dashboard Guide](docs/how-to/runtime-dashboard.md)
 - [Runtime MCP Installation](docs/how-to/runtime-mcp-installation.md)

package/README.zh-CN.md

@@ -46,10 +46,11 @@
 ## 运行时管控一览
 
 - **四信号就绪检查**：在进入实现阶段前执行，与实现评分保持独立
-- **运行时管控循环**：工作流推进与阶段事实保持一致，不允许宿主"乐观继续"
-- **重跑检查点**：修复后沿同一条治理路径回到重试或重新进入
-- **执行闭环证据**：Cursor 与 Claude Code 两侧都保留完整的治理执行记录
-- **事后审计收口**：通过同一条运行时收口路径驱动评分、看板、诊断和训练数据提取
+- **先读 `main-agent-orchestration inspect`**：主 Agent 必须先读取 repo-native authoritative surface，再决定下一条全局分支
+- **按需执行 `dispatch-plan`**：只有 surface 明确需要 materialize packet 时，才生成正式派发计划
+- **子代理只执行 `bounded packet`**：子代理只返回 packet 结果，不负责决定下一条全局执行链
+- **`runAuditorHost` 只负责 post-audit close-out**：审计通过后统一写入评分、看板、诊断和训练数据，然后主 Agent 重新读取 `inspect`
+- **旧 worker / 手工 close-out 口径仅保留为历史证据**：可继续审计追溯，但不再是当前 accepted runtime path
 
 ## 看板与 MCP
 
@@ -164,16 +165,16 @@ npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit uninstall
 
 ### 核心组件
 
-| 组件                        | 说明                                                                  |
-| :-------------------------- | :-------------------------------------------------------------------- |
-| **`_bmad/`**                | 工作流模块、钩子、提示词、路由与宿主侧资产的规范源                    |
-| **`packages/scoring/`**     | 评分引擎、就绪漂移评估、看板投影、诊断输入与训练数据提取              |
-| **`dashboard`**             | 默认运行时可观测层：实时看板、运行时快照、评分投影                    |
-| **`runtime-mcp`**           | 可选的 MCP 工具接口，通过 `--with-mcp` 显式启用                       |
-| **`speckit-workflow`**      | Specify → Plan → GAPS → Tasks → TDD，并带强制审计循环                 |
-| **`bmad-story-assistant`**  | Story 生命周期路径：Create Story → Party Mode → Dev Story → Implement |
-| **`bmad-bug-assistant`**    | Bug 生命周期路径：RCA → Party Mode → BUGFIX → Implement               |
-| **`bmad-standalone-tasks`** | 针对 TASKS 或 BUGFIX 文档的独立受管子代理执行路径                     |
+| 组件                        | 说明                                                                                                               |
+| :-------------------------- | :----------------------------------------------------------------------------------------------------------------- |
+| **`_bmad/`**                | 工作流模块、钩子、提示词、路由与宿主侧资产的规范源                                                                 |
+| **`packages/scoring/`**     | 评分引擎、就绪漂移评估、看板投影、诊断输入与训练数据提取                                                           |
+| **`dashboard`**             | 默认运行时可观测层：实时看板、运行时快照、评分投影                                                                 |
+| **`runtime-mcp`**           | 可选的 MCP 工具接口，通过 `--with-mcp` 显式启用                                                                    |
+| **`speckit-workflow`**      | Specify → Plan → GAPS → Tasks → TDD，并带强制审计循环                                                              |
+| **`bmad-story-assistant`**  | Story 生命周期入口：主 Agent 先读 `inspect`，按需派发 bounded packet，并在 post-audit 后通过 `runAuditorHost` 收口 |
+| **`bmad-bug-assistant`**    | Bug 生命周期路径：RCA → Party Mode → BUGFIX → Implement，但全局 `inspect -> dispatch-plan -> closeout` 主链仍由主 Agent 控制 |
+| **`bmad-standalone-tasks`** | 针对 TASKS 或 BUGFIX 文档的执行仍先经过主 Agent `inspect`，必要时 `dispatch-plan`，再进入 bounded 子代理实施       |
 
 <details>
 <summary><b>查看目录结构</b></summary>
@@ -195,6 +196,7 @@ BMAD-Speckit-SDD-Flow/
 ## 文档入口
 
 - [快速开始](docs/tutorials/getting-started.md)
+- [主 Agent 编排参考](docs/reference/main-agent-orchestration.md)
 - [消费项目安装指南](docs/how-to/consumer-installation.md)
 - [运行时看板指南](docs/how-to/runtime-dashboard.md)
 - [运行时 MCP 安装](docs/how-to/runtime-mcp-installation.md)

package/_bmad/_config/bmads-runtime.yaml

@@ -0,0 +1,193 @@
+schemaVersion: bmads-runtime/v1
+
+contracts:
+  governance: _bmad/_config/orchestration-governance.contract.yaml
+  dispatch: scripts/orchestration-dispatch-contract.ts
+  stageMapping: _bmad/_config/stage-mapping.yaml
+
+entry:
+  canonicalCommand: bmads
+  aliases:
+    - bmad-speckit
+    - $bmads
+    - /bmads
+    - $bmad-speckit
+    - /bmad-speckit
+  advisory:
+    bmadHelpCommand: bmad-help
+    message: Run bmad-help for full BMAD Method workflow guidance.
+
+layers:
+  - id: layer_1
+    name: Product Definition
+    stages:
+      - id: prd
+        governanceStage: prd
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        recommendedWorkflows:
+          - bmad-bmm-create-product-brief
+          - bmad-bmm-create-prd
+          - bmad-bmm-validate-prd
+        evidence:
+          completionMarker: _bmad-output/runtime/context/layer_1-prd.complete.json
+          primaryArtifacts:
+            - _bmad-output/planning-artifacts/{branch}/prd.md
+            - _bmad-output/planning-artifacts/prd.md
+            - _bmad-output/planning-artifacts/product-brief-*.md
+        next:
+          layer: layer_2
+          stage: arch
+
+  - id: layer_2
+    name: Architecture
+    stages:
+      - id: arch
+        governanceStage: arch
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        recommendedWorkflows:
+          - bmad-bmm-create-architecture
+          - bmad-bmm-validate-architecture
+        evidence:
+          primaryArtifacts:
+            - _bmad-output/planning-artifacts/{branch}/architecture.md
+            - _bmad-output/planning-artifacts/architecture.md
+            - docs/architecture/**
+        next:
+          layer: layer_3
+          stage: epics
+
+  - id: layer_3
+    name: Epic And Story Planning
+    stages:
+      - id: epics
+        governanceStage: epics
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        recommendedWorkflows:
+          - bmad-bmm-create-epics-and-stories
+          - bmad-bmm-sprint-planning
+        evidence:
+          primaryArtifacts:
+            - _bmad-output/planning-artifacts/{branch}/epics.md
+            - _bmad-output/planning-artifacts/epics.md
+        next:
+          layer: layer_3
+          stage: story_create
+      - id: story_create
+        governanceStage: story_create
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        recommendedWorkflows:
+          - bmad-bmm-create-epics-and-stories
+          - bmad-bmm-check-implementation-readiness
+          - bmad-bmm-sprint-planning
+          - bmad-bmm-create-story
+        evidence:
+          primaryArtifacts:
+            - _bmad-output/sprint-status.yaml
+            - docs/stories/**
+        next:
+          layer: layer_4
+          stage: specify
+
+  - id: layer_4
+    name: Speckit Implementation
+    stages:
+      - id: specify
+        governanceStage: specify
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        recommendedWorkflows:
+          - speckit-workflow
+        dispatch:
+          taskType: document
+          hosts: [cursor, claude, codex]
+        next:
+          layer: layer_4
+          stage: plan
+      - id: plan
+        governanceStage: plan
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        recommendedWorkflows:
+          - speckit-workflow
+        dispatch:
+          taskType: document
+          hosts: [cursor, claude, codex]
+        next:
+          layer: layer_4
+          stage: gaps
+      - id: gaps
+        governanceStage: gaps
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        recommendedWorkflows:
+          - speckit-workflow
+        dispatch:
+          taskType: audit
+          hosts: [cursor, claude, codex]
+        next:
+          layer: layer_4
+          stage: tasks
+      - id: tasks
+        governanceStage: tasks
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        recommendedWorkflows:
+          - speckit-workflow
+        dispatch:
+          taskType: document
+          hosts: [cursor, claude, codex]
+        next:
+          layer: layer_4
+          stage: implement
+      - id: implement
+        governanceStage: implement
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        recommendedWorkflows:
+          - speckit-workflow
+        dispatch:
+          taskType: implement
+          hosts:
+            - cursor
+            - claude
+            - codex
+        next:
+          layer: layer_5
+          stage: post_audit
+
+  - id: layer_5
+    name: Closeout
+    stages:
+      - id: post_audit
+        governanceStage: post_audit
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        recommendedWorkflows:
+          - bmad-code-reviewer-lifecycle
+        dispatch:
+          taskType: audit
+          hosts: [cursor, claude, codex]
+        next:
+          layer: layer_5
+          stage: pr_review
+      - id: pr_review
+        governanceStage: post_audit
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        gates:
+          required:
+            - pr_topology_closed
+        next:
+          layer: layer_5
+          stage: release_gate
+      - id: release_gate
+        governanceStage: post_audit
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        gates:
+          required:
+            - release_gate
+            - quality_gate
+            - sprint_audit
+        next:
+          layer: layer_5
+          stage: delivery_truth_gate
+      - id: delivery_truth_gate
+        governanceStage: post_audit
+        requiredGovernanceSignalsFrom: orchestration-governance.contract.yaml
+        gates:
+          required:
+            - delivery_truth_gate
+            - long_run_development_evidence

package/_bmad/_config/governance-remediation.yaml

@@ -4,6 +4,32 @@ packetHosts:
   - cursor
   - claude
   - codex
+execution:
+  enabled: true
+  interactiveMode: main-agent
+  fallbackAutonomousMode: false
+  authoritativeHost: cursor
+  fallbackHosts:
+    - claude
+    - codex
+  dispatch:
+    leaseTimeoutSeconds: 900
+    heartbeatIntervalSeconds: 60
+    maxDispatchAttempts: 3
+  execution:
+    timeoutMinutes: 30
+    maxExecutionAttempts: 2
+  rerunGate:
+    required: true
+    autoSchedule: true
+    maxGateRetries: 2
+  escalation:
+    afterDispatchFailures: 3
+    afterExecutionFailures: 2
+    afterGateFailures: 2
+  projections:
+    emitNonAuthoritativePackets: true
+    archivePath: _bmad-output/runtime/governance/archive
 provider:
   mode: openai-compatible
   id: openai-governance

package/_bmad/_config/main-agent-quality-gate.thresholds.json

@@ -0,0 +1,46 @@
+{
+  "version": 1,
+  "gateId": "main-agent-quality-gate",
+  "maxTodoStubs": 0,
+  "maxForbiddenTodoMarkers": 0,
+  "maxMissingKeyPaths": 0,
+  "requiredKeyPaths": [
+    "scripts/main-agent-orchestration.ts",
+    "scripts/main-agent-unified-ingress.ts",
+    "scripts/main-agent-codex-worker-adapter.ts",
+    "scripts/main-agent-release-gate.ts",
+    "scripts/main-agent-delivery-truth-gate.ts",
+    "scripts/main-agent-quality-gate.ts",
+    "scripts/main-agent-chaos-scenarios.ts",
+    "_bmad/_config/orchestration-governance.contract.yaml",
+    "_bmad-output/runtime/governance/user_story_mapping.json",
+    "tests/acceptance/main-agent-release-gate-contract.test.ts",
+    "tests/acceptance/main-agent-delivery-truth-gate.test.ts",
+    "tests/acceptance/governance-rerun-gate-closure.test.ts",
+    "tests/acceptance/main-agent-host-matrix-pr-orchestration.test.ts",
+    "tests/acceptance/sprint-status-authorized-update.test.ts",
+    "docs/ops/main-agent-capability-traceability-matrix.md"
+  ],
+  "requiredAcceptanceTests": [
+    "tests/acceptance/main-agent-quality-gate-thresholds.test.ts",
+    "tests/acceptance/main-agent-release-gate-contract.test.ts",
+    "tests/acceptance/main-agent-delivery-truth-gate.test.ts",
+    "tests/acceptance/governance-rerun-gate-closure.test.ts",
+    "tests/acceptance/main-agent-host-matrix-pr-orchestration.test.ts",
+    "tests/acceptance/main-agent-run-loop-e2e.test.ts",
+    "tests/acceptance/main-agent-codex-agent-parity.test.ts",
+    "tests/acceptance/main-agent-codex-worker-adapter-e2e.test.ts",
+    "tests/acceptance/main-agent-unified-ingress-e2e.test.ts"
+  ],
+  "requiredCodexProofPaths": [
+    "tests/acceptance/main-agent-codex-agent-parity.test.ts",
+    "tests/acceptance/main-agent-codex-worker-adapter-e2e.test.ts",
+    "tests/acceptance/main-agent-codex-consumer-five-layer-e2e.test.ts",
+    "tests/acceptance/reviewer-codex-noop-guard.test.ts",
+    "tests/acceptance/reviewer-rollout-rollback-proof.test.ts"
+  ],
+  "forbiddenTodoMarkers": [
+    "failTodo('main-agent-quality-gate'",
+    "TODO implementation not completed"
+  ]
+}

package/_bmad/_config/orchestration-governance.contract.yaml

@@ -0,0 +1,210 @@
+version: 1
+contract_id: orchestration-governance
+owner: main-agent-orchestration
+status: draft
+updated_at: "2026-04-26T13:23:00+08:00"
+
+description: >
+  Single strategy source for main-agent orchestration governance.
+  Main loop must read this contract for signal policy and stage gate requirements.
+
+sources_of_truth:
+  strategy_contract: "_bmad/_config/orchestration-governance.contract.yaml"
+  runtime_index: "_bmad-output/runtime/governance/user_story_mapping.json"
+  runtime_index_schema: "_bmad-output/runtime/governance/user_story_mapping.schema.json"
+
+consumption_rules:
+  main_loop_must_read:
+    - strategy_contract
+    - runtime_index
+  forbidden_strategy_sources:
+    - ad_hoc_markdown_rules
+    - inline_hardcoded_signal_thresholds
+    - duplicated_yaml_contracts
+
+signals:
+  - id: p0_journey_coverage
+    category: delivery_readiness
+    severity_default: block
+    description: P0 journeys must be covered and mapped to executable task chains.
+    required_evidence:
+      - journey_coverage_matrix
+      - mapped_journey_task_chain
+  - id: smoke_e2e_readiness
+    category: test_readiness
+    severity_default: block
+    description: Critical journeys must have runnable smoke/E2E proof.
+    required_evidence:
+      - smoke_test_entry
+      - smoke_execution_output
+  - id: evidence_proof_chain
+    category: evidence_integrity
+    severity_default: block
+    description: Claims must be backed by command output and persistent artifacts.
+    required_evidence:
+      - command_output_summary
+      - artifact_paths
+  - id: cross_document_traceability
+    category: traceability
+    severity_default: warn
+    description: Requirement/plan/tasks/test references must be mutually traceable.
+    required_evidence:
+      - requirement_to_story_link
+      - story_to_test_link
+  - id: orchestration_governance_completeness
+    category: control_plane_integrity
+    severity_default: warn
+    description: Main loop must execute full control steps without bypass.
+    required_evidence:
+      - inspect_dispatch_ingest_sequence
+      - gates_loop_state_transition
+      - closeout_gate_chain
+
+stage_requirements:
+  prd:
+    required_signals:
+      - p0_journey_coverage
+      - cross_document_traceability
+  arch:
+    required_signals:
+      - p0_journey_coverage
+      - cross_document_traceability
+  epics:
+    required_signals:
+      - p0_journey_coverage
+      - cross_document_traceability
+  story_create:
+    required_signals:
+      - p0_journey_coverage
+      - cross_document_traceability
+  story_audit:
+    required_signals:
+      - evidence_proof_chain
+      - cross_document_traceability
+  specify:
+    required_signals:
+      - p0_journey_coverage
+      - cross_document_traceability
+  plan:
+    required_signals:
+      - p0_journey_coverage
+      - smoke_e2e_readiness
+      - cross_document_traceability
+  gaps:
+    required_signals:
+      - p0_journey_coverage
+      - smoke_e2e_readiness
+      - evidence_proof_chain
+  tasks:
+    required_signals:
+      - p0_journey_coverage
+      - smoke_e2e_readiness
+      - cross_document_traceability
+  implement:
+    required_signals:
+      - p0_journey_coverage
+      - smoke_e2e_readiness
+      - evidence_proof_chain
+      - orchestration_governance_completeness
+    signal_overrides:
+      orchestration_governance_completeness: warn
+  post_audit:
+    required_signals:
+      - evidence_proof_chain
+      - cross_document_traceability
+      - orchestration_governance_completeness
+    signal_overrides:
+      orchestration_governance_completeness: block
+  epic_create:
+    required_signals:
+      - p0_journey_coverage
+      - cross_document_traceability
+  epic_complete:
+    required_signals:
+      - smoke_e2e_readiness
+      - evidence_proof_chain
+      - orchestration_governance_completeness
+    signal_overrides:
+      orchestration_governance_completeness: block
+
+mapping_contract:
+  required_fields:
+    - requirementId
+    - sourceType
+    - epicId
+    - storyId
+    - flow
+    - sprintId
+    - allowedWriteScope
+    - status
+  consistency_rules:
+    - requirement_to_flow_must_exist
+    - requirement_to_story_must_be_unique_per_active_status
+    - changed_files_must_fit_allowed_write_scope
+    - flow_phase_must_match_orchestration_state
+    - sprint_id_must_be_present
+    - reroute_must_deactivate_previous_active_mapping
+    - closeout_must_write_back_mapping_status
+
+gate_policy:
+  missing_required_signal: block
+  missing_required_mapping_field: block
+  mapping_consistency_violation: block
+  warn_threshold:
+    max_warn_signals_before_block: 2
+
+adaptive_intake_governance_gate:
+  gate_name: adaptive_intake_governance_gate
+  applies_to_flows:
+    - story
+    - bugfix
+    - standalone_tasks
+  trigger_points:
+    - on_new_requirement_intake
+    - on_reroute
+    - before_closeout
+  match_scoring:
+    domain_fit: 0.30
+    dependency_fit: 0.25
+    sprint_fit: 0.20
+    risk_fit: 0.15
+    readiness_fit: 0.10
+  decision_thresholds:
+    min_confidence_for_auto_match: 0.70
+    min_confidence_for_warn: 0.55
+  consistency_dimensions:
+    mapping_consistency:
+      required:
+        - requirement_to_story_unique_active
+        - flow_to_story_type_consistent
+        - allowed_write_scope_consistent
+    lifecycle_consistency:
+      required:
+        - orchestration_state_aligned_with_mapping_status
+        - reroute_history_consistent
+        - closeout_result_written_back
+    sprint_consistency:
+      required:
+        - sprint_id_valid
+        - backlog_sync_record_present
+        - capacity_overflow_requires_replan
+  verdict_policy:
+    any_required_check_fail: block
+    warn_if_confidence_below_auto_match: true
+    pass_requires:
+      - mapping_consistency
+      - lifecycle_consistency
+      - sprint_consistency
+
+observability_targets:
+  match_precision_at_1:
+    target: 0.80
+  reroute_recovery_rate:
+    target: 0.90
+  sprint_sync_latency_minutes:
+    target_max: 30
+  mapping_conflict_rate:
+    target_max: 0.05
+  governance_bypass_count:
+    target: 0
+