@xulthekl/team-flow 0.36.4 → 0.38.0

package/.claude/always/phase-guard.md

@@ -1,3 +1,3 @@
-# team-flow v0.36.4 | 阶段: {{state}} | 工作流: {{workflow}}
+# team-flow v0.38.0 | 阶段: {{state}} | 工作流: {{workflow}}
 当前阶段允许的操作由 workflow-start 路由规则定义。
 禁止跨越 DP gate 进入下一阶段。变更范围以 execution-contract.md 的 Intent Lock 为准。

package/.claude-plugin/marketplace.json

@@ -9,7 +9,7 @@
     {
       "name": "team-flow",
       "description": "8-state spec workflow + compound global compounding + architecture-design (4A/DDD) + local HTML prototype + product-level orchestration + bootstrap + e2e + session handoff + workflow feedback. 24 skills + 15 agents with embedded TDD, SDD, code review, debugging, delta spec sync, and design-system-driven prototyping.",
-      "version": "0.36.4",
+      "version": "0.38.0",
       "source": "./",
       "author": {
         "name": "LT",

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "team-flow",
-  "version": "0.36.4",
+  "version": "0.38.0",
   "description": "Unified workflow plugin: team-flow (spec-driven dev: TDD/SDD/code-review/debugging) + compound-engineering core subset (brainstorm/plan/compound/strategy/ideate/proof, global compounding) + architecture-design (4A+DDD incremental design & global compounding) + prototype (local HTML prototype, zero external deps) + e2e (Playwright E2E, AC-driven) + workflow-orchestrator (product-level workflow orchestration) + workflow-bootstrap (existing project onboarding) + session-handoff (context transfer) + workflow-feedback (issue tracking). 24 skills + 15 agents, one install.",
   "source": "./",
   "author": {

package/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "team-flow",
-  "version": "0.36.4",
+  "version": "0.38.0",
   "description": "Spec-first workflow that bridges OpenSpec-style planning and Superpowers-style execution discipline.",
   "author": {
     "name": "MageByte",

package/.cursor-plugin/marketplace.json

@@ -5,7 +5,7 @@
   },
   "metadata": {
     "description": "Unified workflow plugin marketplace for Cursor (team-flow: team-flow + compound + architecture-design + prototype).",
-    "version": "0.36.4"
+    "version": "0.38.0"
   },
   "plugins": [
     {

package/.cursor-plugin/plugin.json

@@ -2,7 +2,7 @@
   "name": "team-flow",
   "displayName": "team-flow",
   "description": "Unified workflow plugin: team-flow (spec-driven dev: TDD/SDD/code-review/debugging) + compound-engineering core subset (brainstorm/plan/compound/strategy/ideate/proof, global compounding) + architecture-design (4A+DDD incremental design & global compounding) + prototype (local HTML prototype, zero external deps) + e2e (Playwright E2E, AC-driven) + workflow-orchestrator (product-level workflow orchestration) + workflow-bootstrap (existing project onboarding) + session-handoff (context transfer) + workflow-feedback (issue tracking). 24 skills + 15 agents, one install.",
-  "version": "0.36.4",
+  "version": "0.38.0",
   "author": {
     "name": "LT",
     "url": "https://github.com/LT"

package/.github/plugin/marketplace.json

@@ -6,13 +6,13 @@
   },
   "metadata": {
     "description": "Unified workflow plugins and skills for AI coding agents (team-flow: team-flow + compound + architecture-design + prototype).",
-    "version": "0.36.4"
+    "version": "0.38.0"
   },
   "plugins": [
     {
       "name": "team-flow",
       "description": "Unified workflow with planning artifacts, execution contracts, TDD, review gates, systematic debugging, delta spec sync, architecture-design, and local HTML prototyping.",
-      "version": "0.36.4",
+      "version": "0.38.0",
       "source": ".",
       "author": {
         "name": "LT",

package/AGENTS.md

@@ -49,6 +49,7 @@ spec 驱动开发过程。8 态变更机：`exploring → specifying → bridgin
 
 ### 4. prototype（1 skill，本地 HTML 原型，v0.15.0 内部编排重构）
 - 全局一份 `prototype/`（与 PRD 同级，git 分支隔离），零外部依赖、可离线
+- **v0.37.0 版本 worktree 隔离**：原型独立仓库（有远程），版本分支 `prd-vN` 的 worktree 由 `tf prototype branch <prd-vN>` 创建/复用（隔离 + 团队拉取），版本收尾 `tf prototype deisolate <prd-vN> --merge --clean` merge 回主干（§68.4）；原型 worktree 布局 `<workspace>/.worktrees/<prd-vN>/prototype/`
 - **内部编排器（v0.15.0）**：主代理只编排不实施——环境探查(`prototype-env-scout`)→方案/人工评审→绘制(`prototype-builder`)→评审(`prototype-reviewer`)→循环→人工路由
 - **v0.20.0 产出质量闸门**：builder→reviewer 间主代理强制 `test -f && test -s` 校验入口产物落盘（不信 status=done 字面值，缺失则 resume，不空转评审）；后台子代理依赖 `<task-notification>` 完成通知，禁反复 `TaskOutput(block=true)` 轮询（设计 §22）
 - **v0.21.0 决策点中继（stop-and-resume）**：子代理（无 AskUserQuestion）决策点经 `SendMessage(to: main)` 发问即停，主代理收 task-notification 代问用户、回传自动 resume 子代理续跑；主代理须维护请求-应答匹配（设计 §22.1.1，修订 §18.1.2，LT 实证确认）

package/CHANGELOG.md

@@ -6,6 +6,43 @@ The format loosely follows Keep a Changelog.
 
 ## [Unreleased]
 
+## [0.38.0] - 2026-08-06
+
+### Added（E2E case 记录 + 复利闭环 + 下游确认）
+
+- **test-matrix 支持 E2E case（`test_tier=e2e`）**：对齐 v0.12 §42.4 三值承诺（unit/integration/e2e），实现补齐设计——test-strategy §3 从"e2e 独立覆盖不计入本矩阵"改为"含 UI 交互/跨层 AC 显式评估 E2E 必要性：纳入或记录排除理由"；金字塔分母 = unit+integration
+- **复利回写 E2E 层级**：`tf test-merge` rewriteIndex 的 `test_tier_breakdown` 增加 e2e 统计（`unit:N, integration:N, e2e:N`）；extractSummary 识别 `E2E cases:` 行
+- **下游 E2E 确认（release-archivist Step 5b）**：检测 test-matrix 含 E2E case → 无 e2e 报告 → AskUserQuestion 确认是否执行 E2E 测试（是 → 引导 `/e2e`；否 → 记录 skip + 理由，不 FAIL）
+- **LLM 消费点 e2e 豁免（防断链）**：release-archivist Step 2b / build-executor implementer / code-reviewer 三个消费点的 coverage 分母与金字塔配比排除 `test_tier=e2e`
+- **生成器/上游同步**：contract-builder Generation Protocol 第 7 步 E2E 评估 + Summary/契约段 E2E 行；test-matrix-export `TEST_KIND_TO_TIER` 加 `playwright_prototype`/`playwright_integration` → e2e；spec-writer Scenario 标签补 `##### E2E:`；design-methods-detail test_tier 表补 e2e；test-quality-rules 规则 11 明确 e2e case 合法落点 `e2e/`
+
+### Bug Fixes（feedback 2026-08-05，test-merge 五 bug）
+
+- **resolveDeferred 误删 Current Cases**：原全文件 regex `\|<caseId>\|[^\n]*\n` 把 baselines Current Cases 表正常 case 行全部误删（表体清空、Evolution Log 计数保留）→ 改为仅删除 `## Deferred Items` 段内被新 case 覆盖的行
+- **extractSummary 遇 markdown bold 返回 0**：regex `Total cases:\s*(\d+)` 不匹配 `**87**`（C1 矩阵 LLM 手写加粗）→ 容忍可选 `**` 前缀
+- **dry-run 实为真写**：dryRun 仅 gitCommit 生效，Step 2-5 照常写盘 → mergeBaselines/resolveDeferred/appendChangelog/rewriteIndex 全链路 dryRun 透传，dry-run 不写任何文件
+- **mergeExistingBaseline `startsWith('')` 恒真**：Current Cases"覆盖式"语义失效，旧 case 行每次 merge 累积 → 修正退出条件，旧行被替换
+- **`.filter(Boolean)` 丢空列错位**：mock 等空列导致 cells 索引左移，work_mode/test_file/test_method_name 静默错位（E2E case 的 mock 列必空，高发）→ 改 `slice(1,-1)` 按索引保留空列
+- **Evolution Log 幂等**：同 change 同日重复执行不重复 append
+
+### Tests
+
+- **tests/lib/test-merge.test.mjs（新增）**：test-merge 是全库唯一无测试覆盖的核心 merge 命令，补齐 11 个用例（resolveDeferred 不误删 / extractSummary bold / 覆盖式 + 幂等 / 空列列位 / dry-run 不写盘 / E2E 复利统计）
+
+### 来源
+
+- workflow-feedback 2026-08-05（P1-1~P1-4 + P2-1）交叉确认：2 个独立 Explore agent 制品链核查 + resolveDeferred 实证（旧代码表体清空 vs 新代码保留）
+
+## [0.37.0] - 2026-08-05
+
+### Added（阶段同步 + 原型版本隔离，设计增强方案 v0.15 §68）
+
+- **CLI 新命令 `tf publish`**：阶段产物白名单提交 + 可选推送（`--prd / --arch / --changes <dir> / --all` + `--push` 确认制 + `--dry-run`）；复用 arch-merge 白名单 commit 模式 + 外部脏文件检测 + 独立写锁（`.publish.lock`）
+- **CLI 新命令 `tf prototype branch` / `tf prototype deisolate`**：原型仓库（独立 + 远程）PRD 版本 worktree 隔离——`branch <prd-vN>` 创建/复用版本 worktree（隔离 + 团队一键拉取）；`deisolate <prd-vN> --merge --clean` 版本收尾 merge 回主干 + 清理（方案 A：merge 时机 = 版本收尾）
+- **5 个同步门禁点（G1-G5，§68.2）**：S2 冻结 / ARCH 完成 / orchestrator 结束 / execution-contract 完成 / change closing——各点 AskUserQuestion 三选项（提交并推送 / 仅提交 / 暂不同步），团队协作阶段产物同步
+- **制品链路径版本化修正（§68.5）**：原型引用 = 语义相对 + 版本绑定；spec-writer / build-executor / e2e / contract-builder 声明原型版本 worktree 路径；`prototype-sync --prototype-dir` 缺省按版本解析 worktree（`derivePrdVersion`/`resolvePrototypeDir`，change-brief `upstream_plan_ref` 推导，worktree 缺失回退主仓库并提示）
+- **文档**：设计增强方案 v0.15（§68 阶段同步 + 原型版本隔离 + §69 决策落实状态）
+
 ## [0.36.4] - 2026-08-05
 
 ### Added（Wave 审查纪律 + 上下文轮转）

package/GEMINI.md

@@ -8,7 +8,7 @@ The workflow is self-contained and does not require OpenSpec or Superpowers at r
 
 
 <!-- team-flow-phase-guard-start -->
-# team-flow v0.36.4 | 阶段: {{state}} | 工作流: {{workflow}}
+# team-flow v0.38.0 | 阶段: {{state}} | 工作流: {{workflow}}
 当前阶段允许的操作由 workflow-start 路由规则定义。
 禁止跨越 DP gate 进入下一阶段。变更范围以 execution-contract.md 的 Intent Lock 为准。
 <!-- team-flow-phase-guard-end -->

package/INSTALL.md

@@ -7,7 +7,7 @@
 - [Fission-AI/OpenSpec](https://github.com/Fission-AI/OpenSpec) — 规划引擎（Schema 验证、Delta Spec、工件解析）
 - [obra/superpowers](https://github.com/obra/superpowers) — 执行纪律（TDD 铁律、SDD、系统化调试、代码审查）
 
-当前发布版本：**v0.36.4**。
+当前发布版本：**v0.38.0**。
 
 ---
 

package/README.md

@@ -1,6 +1,6 @@
 # team-flow
 
-> 当前版本：`v0.36.4`
+> 当前版本：`v0.38.0`
 
 > 统一插件：**team-flow**（spec 驱动开发）+ **compound-engineering 核心子集**（全局复利）+ **architecture-design**（4A+DDD 增量设计）+ **prototype**（本地 HTML 原型）+ **e2e**（AC 驱动 E2E）+ **workflow-orchestrator**（产品级编排）+ **workflow-bootstrap**（既有项目接入）。一次安装，七套能力协同。
 

package/agents/architecture-reviewer.md

@@ -1,17 +1,21 @@
 ---
 name: architecture-reviewer
-description: 架构设计自动审查 agent——独立上下文、只读审查，对照 PRD/plan/change-brief + 全局基线 + conventions 审查架构产出质量（6 维度 A1-A6），填补 architecture-design 阶段"设计完无审查"的缺口。循环修正 ≤3 轮 + 收敛检测。
+description: 架构设计自动审查 agent——独立上下文，对照 PRD/plan/change-brief + 全局基线 + conventions 审查架构产出质量（6 维度 A1-A6），填补 architecture-design 阶段"设计完无审查"的缺口。循环修正 ≤3 轮 + 收敛检测。直接写审查报告文件，返回摘要 + 文件路径。
 
 model: inherit
 color: yellow
-tools: ["Read", "Bash", "Grep", "Glob"]
+tools: ["Read", "Bash", "Grep", "Glob", "Write"]
 ---
 
-You are an independent Architecture Consistency Reviewer. You check whether architecture design deliverables are structurally complete, consistent with the PRD/plan/change-brief, aligned with the global architecture baseline, and compliant with project conventions. You produce a graded review report across 6 dimensions. You NEVER modify any file — your role is strictly read-only analysis.
+You are an independent Architecture Consistency Reviewer. You check whether architecture design deliverables are structurally complete, consistent with the PRD/plan/change-brief, aligned with the global architecture baseline, and compliant with project conventions. You produce a graded review report across 6 dimensions and write it to disk.
 
 ## The Iron Law
 
-You are an independent reviewer. You did NOT participate in producing the architecture design. You read the deliverables fresh, compare them against the baseline and requirements, and report findings. You must NEVER write, edit, create, or delete any file. Your report is delivered in your response — the orchestration layer persists it to `changes/<name>/architecture/auto-review.md`（change 模式）或 `docs/architecture/iterations/vN/auto-review.md`（product 模式，v0.36.0）.
+You are an independent reviewer. You did NOT participate in producing the architecture design. You read the deliverables fresh, compare them against the baseline and requirements, and report findings. You MUST write the review report to disk:
+- **change 模式**：`changes/<name>/architecture/auto-review.md`
+- **product 模式**：`docs/architecture/iterations/vN/auto-review.md`
+
+After writing, return a summary of the verdict + the report file path. Do NOT modify any architecture deliverable files — only the review report.
 
 ## Inputs
 
@@ -149,7 +153,7 @@ Aggregate all findings and determine verdict.
 - **Architecture Dir**: {architecture_dir}
 - **Global Arch Dir**: {global_arch_dir}
 - **Review round**: {N}
-- **Reviewer**: architecture-reviewer agent (independent, read-only)
+- **Reviewer**: architecture-reviewer agent (independent)
 - **Timestamp**: {ISO 8601}
 
 ## Verdict: {PASS | PASS_WITH_WARNINGS | FAIL}
@@ -214,6 +218,11 @@ Aggregate all findings and determine verdict.
 | Bounded Contexts | {n} | {m} | {percent}% |
 | APIs | {n} | {m} | {percent}% |
 | DB Changes | {n} | {m} | {percent}% |
+
+## Report File
+
+- **Path**: {report_file_path}
+- **Written**: ✅
 ```
 
 ## Red Lines
@@ -225,9 +234,10 @@ Aggregate all findings and determine verdict.
 - Load and check conventions when conventions_config is provided
 - Build an explicit coverage matrix for A4, not a vibes-based assessment
 - Give a clear, unambiguous verdict
+- Write the review report to the specified path, then return summary + file path
 
 **DON'T:**
-- Modify any file — you are read-only, report goes in your response
+- Modify any architecture deliverable files — only write the review report
 - Guess or infer requirements not explicitly in change-brief or plan
 - FAIL on Important findings — FAIL is Critical-only
 - Skip the Pre-check phase and jump straight to semantic analysis

package/agents/change-split-auditor.md

@@ -1,17 +1,17 @@
 ---
 name: change-split-auditor
-description: plan.md 拆分质量审计。在 ce-plan 产出 plan.md 后、change 脚手架前，独立审计拆分质量——PRD 需求覆盖矩阵、DAG 无环、粒度均衡、字段完整性、所有权粒度合规（v0.9 根因判据硬门禁：切碎检测/无设计单元/跨单元臃肿）。输出 PASS/FAIL verdict。只读审查。
+description: plan.md 拆分质量审计。在 ce-plan 产出 plan.md 后、change 脚手架前，独立审计拆分质量——PRD 需求覆盖矩阵、DAG 无环、粒度均衡、字段完整性、所有权粒度合规（v0.9 根因判据硬门禁：切碎检测/无设计单元/跨单元臃肿）。输出 PASS/FAIL verdict + 审查报告文件。
 
 model: inherit
 color: cyan
-tools: ["Read", "Bash", "Grep", "Glob"]
+tools: ["Read", "Bash", "Grep", "Glob", "Write"]
 ---
 
-You are an independent Change Split Auditor. You audit the quality of a plan.md's change decomposition BEFORE any change directories are created. You check 5 dimensions: requirement coverage, DAG validity, granularity balance, field completeness, and ownership granularity compliance (v0.9 root-cause criterion). You NEVER modify any file — your role is strictly read-only analysis.
+You are an independent Change Split Auditor. You audit the quality of a plan.md's change decomposition BEFORE any change directories are created. You check 5 dimensions: requirement coverage, DAG validity, granularity balance, field completeness, and ownership granularity compliance (v0.9 root-cause criterion). You write the audit report to disk and return a summary.
 
 ## The Iron Law
 
-You are a read-only auditor. You did NOT produce the plan or the PRD. You read both artifacts, cross-reference them, and produce a verdict with findings. You must NEVER write, edit, create, or delete any file. On FAIL, the orchestration layer carries your findings back to ce-plan for deepening.
+You are an auditor. You did NOT produce the plan or the PRD. You read both artifacts, cross-reference them, and produce a verdict with findings. You MUST write the audit report to disk at the path specified by the orchestration layer (typically `prd/vN/change-split-audit.md`). On FAIL, the orchestration layer carries your findings back to ce-plan for deepening. Do NOT modify plan.md or PRD — only write the audit report.
 
 ## Inputs
 
@@ -139,7 +139,7 @@ FAIL → orchestration layer returns findings to ce-plan for deepening, then re-
 - **PRD**: {prd_path}
 - **Changes found**: {n}
 - **PRD requirements found**: {m}
-- **Auditor**: change-split-auditor agent (independent, read-only)
+- **Auditor**: change-split-auditor agent (independent)
 
 ## Verdict: {PASS | FAIL}
 
@@ -193,6 +193,11 @@ Coverage: {covered}/{total} ({percent}%)
 | 1 | D1 | — | Gap | Critical | PRD §7.3 报表导出未被任何 change 覆盖 | 新增 change 或扩展现有 change scope |
 | 2 | D2 | change-1,3 | Cycle | Critical | change-1→change-3→change-1 循环依赖 | 重新划分边界 |
 | ... | | | | | | |
+
+## Report File
+
+- **Path**: {report_file_path}
+- **Written**: ✅
 ```
 
 ## Red Lines
@@ -211,7 +216,7 @@ Coverage: {covered}/{total} ({percent}%)
 - Recognize legitimate base/infra changes: a change that holds **complete** ownership of a layer/contract (e.g. "read-model CQRS query base") is self-contained → PASS, even if upper changes consume it
 
 **DON'T:**
-- Modify plan.md or PRD — you are read-only
+- Modify plan.md or PRD — only write the audit report
 - FAIL on numerical granularity alone (D3, advisory); but DO FAIL on ownership violation (D5, Critical)
 - Judge D5 by scope **text labels** alone (e.g. "looks like a module") — judge by **ownership analysis**: does the architecture-design of this change overlap with another's?
 - Ignore implicit dependencies — flag them as Important even if they might be intentional

package/agents/code-reviewer.md

@@ -1,21 +1,21 @@
 ---
 name: code-reviewer
-description: 代码审查 agent——审查代码质量、规格合规性、架构合理性、实现完整性。在 build-executor 执行后、release-archivist 归档前，独立审查实现质量。只读审查，不修改任何文件。
+description: 代码审查 agent——审查代码质量、规格合规性、架构合理性、实现完整性。在 build-executor 执行后、release-archivist 归档前，独立审查实现质量。直接写审查报告文件，返回摘要 + 文件路径。
 
 model: inherit
 color: blue
-tools: ["Read", "Grep", "Glob", "Bash"]
+tools: ["Read", "Grep", "Glob", "Bash", "Write"]
 skills:
   - code-reviewer
 ---
 
-You are an independent Code Reviewer. You review code changes for quality, spec compliance, architecture soundness, and implementation completeness. You NEVER modify any file — your role is strictly read-only analysis.
+You are an independent Code Reviewer. You review code changes for quality, spec compliance, architecture soundness, and implementation completeness. You write the review report to disk and return a summary.
 
 **Your preloaded Skill contains the detailed methodology** (6-step review process, calibration rules, severity definitions, critical rules). Follow it for HOW. This prompt defines WHO you are and WHAT you must deliver.
 
 ## Iron Law
 
-You are a read-only reviewer. You did NOT write the code. You read the implementation, compare it against specs/requirements, check code quality, and produce a verdict with findings. You must NEVER write, edit, create, or delete any file. Your review report is delivered in your response.
+You are a reviewer. You did NOT write the code. You read the implementation, compare it against specs/requirements, check code quality, and produce a verdict with findings. You MUST write the review report to disk at the path specified by the orchestration layer (typically `changes/<name>/code-review.md`). Do NOT modify any implementation files — only write the review report.
 
 ## Inputs
 
@@ -37,8 +37,9 @@ Produce a structured review report with the following sections:
 3. **Findings table**: severity, category, file, line, description, suggestion
 4. **Spec compliance matrix**: spec requirement → implementation status
 5. **Summary**: key strengths, critical issues, recommendations
+6. **Report File**: path to the written report file + confirmation
 
-Your preloaded Skill provides the full report template and formatting guidelines.
+Your preloaded Skill provides the full report template and formatting guidelines. Write the report to disk, then return summary + file path.
 
 ## Verdict Criteria
 
@@ -57,7 +58,7 @@ Your preloaded Skill defines severity levels (Critical / Important / Minor) and
 - Give a clear, unambiguous verdict
 
 **DON'T:**
-- Modify any file — you are read-only
+- Modify any implementation files — only write the review report
 - Guess at spec requirements not explicitly stated
 - Skip the spec compliance check (it's the core deliverable)
 - Mark style issues as Critical (they're Minor)

package/agents/cross-change-consistency-checker.md

@@ -1,18 +1,18 @@
 ---
 name: cross-change-consistency-checker
-description: 跨 change 冲突检测。在多 change 并行执行期或 change closing 时，检测共享聚合/实体被多个 change 修改、API 签名变更影响、与全局架构锚点漂移、与原型 testid 契约漂移。只读审查。
+description: 跨 change 冲突检测。在多 change 并行执行期或 change closing 时，检测共享聚合/实体被多个 change 修改、API 签名变更影响、与全局架构锚点漂移、与原型 testid 契约漂移。直接写冲突报告文件，返回摘要 + 文件路径。
 
 
 model: inherit
 color: red
-tools: ["Read", "Bash", "Grep", "Glob"]
+tools: ["Read", "Bash", "Grep", "Glob", "Write"]
 ---
 
-You are an independent Cross-Change Consistency Checker. You detect conflicts across multiple parallel or sequential changes: shared aggregate mutations, API signature breakage, architecture anchor drift, and prototype testid contract drift. You NEVER modify any file — your role is strictly read-only analysis.
+You are an independent Cross-Change Consistency Checker. You detect conflicts across multiple parallel or sequential changes: shared aggregate mutations, API signature breakage, architecture anchor drift, and prototype testid contract drift. You write the conflict report to disk and return a summary.
 
 ## The Iron Law
 
-You are a read-only checker. You did NOT implement any change. You read all change artifacts, global architecture anchors, and the prototype, then report conflicts. You must NEVER write, edit, create, or delete any file. Your conflict report is delivered in your response.
+You are a checker. You did NOT implement any change. You read all change artifacts, global architecture anchors, and the prototype, then report conflicts. You MUST write the conflict report to disk at the path specified by the orchestration layer (typically `changes/<name>/cross-change-consistency.md` or a global report path). Do NOT modify any change artifacts — only write the conflict report.
 
 ## Inputs
 
@@ -106,7 +106,7 @@ Use `grep` to search for API references across change specs and design docs.
 - **Changes analyzed**: {change-1, change-2, ...}
 - **Architecture anchors**: {arch_path}
 - **Prototype**: {prototype_path}
-- **Checker**: cross-change-consistency-checker agent (independent, read-only)
+- **Checker**: cross-change-consistency-checker agent (independent)
 
 ## Result: {CLEAN | CONFLICTS_FOUND}
 
@@ -148,6 +148,11 @@ Use `grep` to search for API references across change specs and design docs.
 | 1 | D1 | change-1, change-3 | OrderAggregate.java | Critical | 合并两个 change 的聚合修改，或拆分为顺序执行 |
 | 2 | D2 | change-1 → change-2 | POST /api/orders | Critical | change-1 需保持向后兼容，或 change-2 同步适配 |
 | ... | | | | | |
+
+## Report File
+
+- **Path**: {report_file_path}
+- **Written**: ✅
 ```
 
 If no conflicts are found, output:
@@ -168,7 +173,7 @@ All 4 dimensions checked. No cross-change conflicts detected.
 - Skip dimensions gracefully when inputs are missing (note the skip, don't crash)
 
 **DON'T:**
-- Modify any change's files — you are read-only
+- Modify any change's files — only write the conflict report
 - Report single-change issues as cross-change conflicts (use code-reviewer for that)
 - Flag shared read-only references as conflicts (two changes READING the same file is fine)
 - Ignore architecture drift because "arch-merge will handle it later" — report it, let the orchestrator decide

package/agents/prd-completeness-reviewer.md

@@ -1,17 +1,17 @@
 ---
 name: prd-completeness-reviewer
-description: PRD 完整性自动评审。在 PRD 写入后、冻结前，独立检查"需求描述是否完整到能支撑后续 plan/spec 实施"（5 维度），输出分级完整性报告。只读审查，不修改任何文件。与 claim verifier 分工——claim verifier 管"说得对不对"（核查事实声明），本 agent 管"说得全不全、能否落地"。
+description: PRD 完整性自动评审。在 PRD 写入后、冻结前，独立检查"需求描述是否完整到能支撑后续 plan/spec 实施"（5 维度），输出分级完整性报告。直接写审查报告文件，返回摘要 + 文件路径。与 claim verifier 分工——claim verifier 管"说得对不对"（核查事实声明），本 agent 管"说得全不全、能否落地"。
 
 model: inherit
 color: yellow
-tools: ["Read", "Bash", "Grep", "Glob"]
+tools: ["Read", "Bash", "Grep", "Glob", "Write"]
 ---
 
-You are an independent PRD Completeness Reviewer. You check whether a PRD is **complete enough to support downstream plan/spec/build work**. You produce a graded completeness report across 5 dimensions. You NEVER modify any file — your role is strictly read-only analysis.
+You are an independent PRD Completeness Reviewer. You check whether a PRD is **complete enough to support downstream plan/spec/build work**. You produce a graded completeness report across 5 dimensions and write it to disk.
 
 ## The Iron Law
 
-You are an independent reviewer. You did NOT participate in producing the PRD. You read the PRD (and CONCEPTS.md if present) fresh and assess whether the requirement description is complete enough to be actionable. You must NEVER write, edit, create, or delete any file. Your report is delivered in your response — the orchestration layer persists it to `prd/vN/prd-completeness-review.md`.
+You are an independent reviewer. You did NOT participate in producing the PRD. You read the PRD (and CONCEPTS.md if present) fresh and assess whether the requirement description is complete enough to be actionable. You MUST write the review report to disk at `prd/vN/prd-completeness-review.md`. Do NOT modify the PRD or any other artifact — only write the review report.
 
 ## Division of Labor with claim verifier
 
@@ -94,7 +94,7 @@ Aggregate findings → final verdict per the Judgment Criteria.
 - **PRD**: {prd_path}
 - **CONCEPTS**: {concepts_path or "N/A"}
 - **Review round**: {N}
-- **Reviewer**: prd-completeness-reviewer agent (independent, read-only)
+- **Reviewer**: prd-completeness-reviewer agent (independent)
 - **Timestamp**: {ISO 8601}
 
 ## Verdict: {PASS | PASS_WITH_WARNINGS | FAIL}
@@ -131,6 +131,11 @@ Aggregate findings → final verdict per the Judgment Criteria.
 ## Completeness Statistics
 - Requirements with complete user story (D1): {m}/{n} ({percent}%)
 - Requirements with verifiable AC (D2): {m}/{n} ({percent}%)
+
+## Report File
+
+- **Path**: {report_file_path}
+- **Written**: ✅
 ```
 
 ## Red Lines
@@ -143,7 +148,7 @@ Aggregate findings → final verdict per the Judgment Criteria.
 - Treat vague AC ("表现良好"/"快速响应" without threshold) as a D2 Critical
 
 **DON'T:**
-- Modify any file — you are read-only, report goes in your response
+- Modify the PRD or any other artifact — only write the review report
 - Duplicate claim verification — do NOT fact-check external claims, that is claim verifier's job
 - Infer requirement content the PRD does not state — if the PRD doesn't define it and the dimension requires it, that is a finding, not something to fill in
 - Trigger FAIL on Important or Minor findings — FAIL is Critical-only

package/agents/prototype-reviewer.md

@@ -1,17 +1,17 @@
 ---
 name: prototype-reviewer
-description: 原型 vs PRD 一致性自动评审。在原型产出后、人工评审前，独立检查原型与 PRD 的结构化一致性（6 维度），输出分级一致性报告。只读审查，不修改任何文件。
+description: 原型 vs PRD 一致性自动评审。在原型产出后、人工评审前，独立检查原型与 PRD 的结构化一致性（6 维度），输出分级一致性报告。直接写审查报告文件，返回摘要 + 文件路径。
 
 model: inherit
 color: green
-tools: ["Read", "Bash", "Grep", "Glob"]
+tools: ["Read", "Bash", "Grep", "Glob", "Write"]
 ---
 
-You are an independent Prototype Consistency Reviewer. You check whether an HTML prototype is structurally and semantically consistent with its PRD. You produce a graded consistency report across 6 dimensions. You NEVER modify any file — your role is strictly read-only analysis.
+You are an independent Prototype Consistency Reviewer. You check whether an HTML prototype is structurally and semantically consistent with its PRD. You produce a graded consistency report across 6 dimensions and write it to disk.
 
 ## The Iron Law
 
-You are an independent reviewer. You did NOT participate in producing the PRD or the prototype. You read both artifacts fresh, compare them structurally, and report findings. You must NEVER write, edit, create, or delete any file. Your report is delivered in your response — the orchestration layer persists it to `prd/vN/prototype-auto-review.md`.
+You are an independent reviewer. You did NOT participate in producing the PRD or the prototype. You read both artifacts fresh, compare them structurally, and report findings. You MUST write the review report to disk at `prd/vN/prototype-auto-review.md`. Do NOT modify the prototype or PRD — only write the review report.
 
 ## Inputs
 
@@ -108,7 +108,7 @@ If either path is missing or unreadable, report `FAIL` with reason `INPUT_ERROR`
 - **PRD**: {prd_path}
 - **Prototype**: {prototype_path}
 - **Review round**: {N}
-- **Reviewer**: prototype-reviewer agent (independent, read-only)
+- **Reviewer**: prototype-reviewer agent (independent)
 - **Timestamp**: {ISO 8601}
 
 ## Verdict: {PASS | PASS_WITH_WARNINGS | FAIL}
@@ -166,6 +166,11 @@ If either path is missing or unreadable, report `FAIL` with reason `INPUT_ERROR`
 - Functions: {m}/{n} ({percent}%)
 - Fields: {m}/{n} ({percent}%)
 - Routes: {m}/{n} ({percent}%)
+
+## Report File
+
+- **Path**: {report_file_path}
+- **Written**: ✅
 ```
 
 ## Red Lines
@@ -178,7 +183,7 @@ If either path is missing or unreadable, report `FAIL` with reason `INPUT_ERROR`
 - Distinguish Pre-check (mechanical) from Deep-check (semantic, advisory) findings
 
 **DON'T:**
-- Modify any file — you are read-only, report goes in your response
+- Modify the prototype or PRD — only write the review report
 - Guess or infer PRD content that is not explicitly written — if PRD doesn't define it, it's not a finding
 - Mark style/aesthetics/UX issues — those are for human review only
 - Trigger FAIL on Important findings — FAIL is Critical-only

package/docs/README_en.md

@@ -126,7 +126,7 @@ npm install -g team-flow
 
 ### Version
 
-- Current: `v0.36.4`
+- Current: `v0.38.0`
 - v0.9.1 highlights: DP-4 execution-mode recommendations, a portable runtime across 17 platforms, and a raw-package smoke with no plugin-root variable.
 - Self-contained — no OpenSpec or Superpowers runtime required
 - Upstream: [Fission-AI/OpenSpec](https://github.com/Fission-AI/OpenSpec), [obra/superpowers](https://github.com/obra/superpowers)

package/docs/solutions/INDEX.md

@@ -2,6 +2,7 @@
 <!-- 每条一行，按 severity 降序，≤150 行硬上限 -->
 | date | phase | domain | type | severity | summary | file |
 |------|-------|--------|------|----------|---------|------|
+| 2026-08-06 | cross-phase | general | insight | medium | (no summary) | cross-phase/2026-08-06-no-summary.md |
 | 2026-08-05 | cross-phase | general | insight | medium | (no summary) | cross-phase/2026-08-05-no-summary.md |
 | 2026-08-04 | cross-phase | general | insight | medium | (no summary) | cross-phase/2026-08-04-no-summary.md |
 | 2026-08-03 | cross-phase | general | insight | medium | (no summary) | cross-phase/2026-08-03-no-summary.md |

package/docs/solutions/cross-phase/2026-08-06-no-summary.md

@@ -0,0 +1,17 @@
+---
+phase: cross-phase
+domain: general
+type: insight
+severity: medium
+date: 2026-08-06
+source: 
+---
+
+## 问题描述
+(no summary)
+
+## 根因/模式
+（待补充）
+
+## 预防措施/应用方式
+（待补充）

package/gemini-extension.json

@@ -1,6 +1,6 @@
 {
   "name": "team-flow",
   "description": "Unified workflow plugin: team-flow (spec-driven dev) + compound-engineering core subset + architecture-design (4A/DDD) + prototype (local HTML). 24 skills, one install.",
-  "version": "0.36.4",
+  "version": "0.38.0",
   "contextFileName": "GEMINI.md"
 }

package/hooks/session-start

@@ -1,11 +1,11 @@
 #!/usr/bin/env bash
-# v0.36.4: auto-sync CLI version with plugin version
+# v0.38.0: auto-sync CLI version with plugin version
 set -e
 
 # ═══════════════════════════════════════════════════════════════
 # Plugin version (update this when releasing new versions)
 # ═══════════════════════════════════════════════════════════════
-PLUGIN_VERSION="0.36.4"
+PLUGIN_VERSION="0.38.0"
 
 # ═══════════════════════════════════════════════════════════════
 # Step 1: Auto-sync CLI version with plugin version

package/llms.txt

@@ -3,7 +3,7 @@
 ## Overview
 spec-superflow is a self-contained workflow integration plugin for Claude Code, Cursor, OpenAI Codex CLI/App, GitHub Copilot CLI, Gemini CLI, OpenCode, WorkBuddy, and Trae. It merges spec-driven planning artifacts (proposal, specs, design, tasks) with disciplined execution guardrails (TDD, review gates, controlled handoff) into one unified workflow.
 
-Current version: v0.36.4.
+Current version: v0.38.0.
 
 ## Key Documents
 - README.md: Chinese homepage with full usage guide and FAQ

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xulthekl/team-flow",
-  "version": "0.36.4",
+  "version": "0.38.0",
   "description": "Unified plugin (24 skills + 15 agents) integrating team-flow, compound-engineering, architecture-design, prototype, design-system, workflow-orchestrator, workflow-bootstrap, e2e, session-handoff, workflow-feedback for multi-agent coding tools.",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",

package/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "team-flow",
-  "version": "0.36.4",
+  "version": "0.38.0",
   "description": "Unified workflow plugin: team-flow (spec-driven dev: TDD/SDD/code-review/debugging) + compound-engineering core subset (brainstorm/plan/compound/strategy/ideate/proof, global compounding) + architecture-design (4A+DDD incremental design & global compounding) + prototype (local HTML prototype, zero external deps) + e2e (Playwright E2E, AC-driven) + workflow-orchestrator (product-level workflow orchestration) + workflow-bootstrap (existing project onboarding) + session-handoff (context transfer) + workflow-feedback (issue tracking). 24 skills + 15 agents, one install.",
   "author": {
     "name": "LT"