intentdna 1.5.7 → 1.5.9

package/dist/templates/tdd-strict.dna.yaml

@@ -52,19 +52,103 @@ genes:
 contexts: {}
 
 roles:
+  test_writer:
+    description: Writes failing tests before implementation. Can create test files.
+    tool_permissions:
+      allow: [Read, Edit, Write, Grep, Glob, Bash]
+    scope:
+      read: ["**/*"]
+      write: ["test/**", "tests/**", "spec/**", "__tests__/**"]
+    instructions:
+      - Write the test FIRST — it must fail initially (red phase)
+      - Test should describe the desired behavior, not the implementation
+      - One test case per behavior
+      - Run the test to confirm it fails for the right reason
+      - Do NOT write implementation code
+
   implementer:
-    description: TDD implementer — writes tests first, then code
+    description: Writes minimal code to make tests pass.
     tool_permissions:
       allow: [Read, Edit, Write, Grep, Glob, Bash]
     scope:
       read: ["**/*"]
-      write: ["src/**", "test/**"]
+      write: ["src/**", "lib/**"]
     instructions:
-      - Always write the failing test FIRST
-      - Run the test to confirm it fails (red)
-      - Write the minimum code to make it pass (green)
-      - Refactor if needed, keeping tests green
-      - Never commit with failing tests
-    post_checks:
-      - vitest_run
-      - tsc_no_errors
+      - Write the MINIMUM code to make the failing test pass (green phase)
+      - Do not add extra functionality beyond what the test requires
+      - Do not optimize or clean up code yet
+      - Run the test after each change to verify it passes
+      - Do not modify test files
+
+  refactorer:
+    description: Cleans up code while keeping tests green.
+    tool_permissions:
+      allow: [Read, Edit, Write, Grep, Glob, Bash]
+    scope:
+      read: ["**/*"]
+      write: ["src/**", "lib/**"]
+    instructions:
+      - Clean up the implementation (refactor phase)
+      - Run ALL tests after each refactoring step
+      - If any test breaks, revert immediately
+      - Focus on readability, removing duplication, improving naming
+      - Do NOT add new functionality during refactoring
+      - Do NOT modify test files
+
+workflows:
+  tdd-cycle:
+    name: TDD Cycle
+    description: "Red-Green-Refactor: write failing test, implement minimally, clean up"
+    steps:
+      - id: write_test
+        role: test_writer
+        description: "Write a failing test for the desired behavior (RED)."
+        prompt: |
+          TDD Red Phase:
+          1. Understand the behavior to implement
+          2. Write ONE test that describes this behavior
+          3. Run the test — it MUST fail
+          4. Verify it fails for the right reason (not a syntax error)
+          5. Commit: "test: add failing test for [behavior]"
+
+          Do NOT write any implementation code.
+
+      - id: implement
+        role: implementer
+        depends_on: [write_test]
+        description: "Write minimal code to pass the test (GREEN)."
+        prompt: |
+          TDD Green Phase:
+          1. Read the failing test carefully
+          2. Write the MINIMUM code to make it pass
+          3. Run the test — it must pass now
+          4. Run the full test suite — no regressions
+          5. Commit: "feat: implement [behavior] — test green"
+
+          Rules:
+          - Do NOT add code beyond what the test requires
+          - Do NOT optimize yet
+          - Do NOT modify test files
+
+      - id: refactor
+        role: refactorer
+        depends_on: [implement]
+        description: "Clean up code while keeping all tests green (REFACTOR)."
+        prompt: |
+          TDD Refactor Phase:
+          1. Review the implementation for code smells
+          2. Apply ONE refactoring at a time
+          3. Run ALL tests after each change
+          4. If any test breaks: REVERT immediately
+          5. Commit: "refactor: [what was cleaned up]"
+
+          Focus on:
+          - Remove duplication
+          - Improve naming
+          - Simplify logic
+          - Extract methods/functions if needed
+
+          Do NOT:
+          - Add new features
+          - Modify test files
+          - Change behavior

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "intentdna",
-  "version": "1.5.7",
+  "version": "1.5.9",
   "description": "Intent DNA — Declarative policy layer for AI agent behavior",
   "type": "module",
   "main": "dist/index.js",

package/spec/flutter-rewrite-template-optimization.md

@@ -1,5 +1,15 @@
 # Spec: flutter-rewrite 模板优化
 
+## 最终目标
+
+**把 v1 (GetX) 重写为 v2 (Riverpod)，行为完全一致。**
+
+- behavior-lock: 写测试 = 定义"v1 的行为是什么"
+- rescue: 写代码让测试通过 = "v2 的行为和 v1 一样"
+- 完成标志: 所有测试 green = 重写成功，行为完全一致
+
+所有设计决策必须服务于这个目标。timeout 是安全网（防卡死），不是解法。挂死的测试意味着行为没被验证——rescue 必须修复它们，不能跳过。
+
 ## 问题
 
 flutter-rewrite 模板的 behavior-lock 和 rescue 工作流都有设计缺陷，导致：
@@ -47,14 +57,19 @@ commit: "behavior-lock($MODULE): N tests (X green, Y red from behavior change)"
 3. 增量更新
 4. 输出 diff 摘要
 
-### 禁止的模式
+### 超时策略
 
-| 禁止 | 原因 | 替代 |
-|------|------|------|
-| `sleep N && check` 轮询 | 浪费时间 | 前台跑 |
-| `timeout Nm flutter test` | 测试可能需要很久 | 不设超时 |
-| 全量重写已有测试 | 丢失 rescue 成果 | 增量更新 |
-| 首次跑 `flutter test` | 没实现，编译不过 | `flutter analyze` |
+| 操作 | 允许？ | 原因 |
+|------|--------|------|
+| `flutter test --timeout 30s` | **允许** | per-test 安全网，挂死标 fail 继续下一个 |
+| `timeout Nm flutter test` | 禁止 | 外部杀进程，丢失所有结果 |
+| `sleep N && check` 轮询 | 禁止 | 浪费时间 |
+| 全量重写已有测试 | 禁止 | 丢失 rescue 成果 |
+| 首次跑 `flutter test` | 禁止 | 没实现，编译不过，用 `flutter analyze` |
+
+**挂死的测试分类**：标记为 `hung: needs mock infrastructure`，不是 `failed`。区别：
+- `failed` = 行为不一致，rescue 修 v2 代码
+- `hung` = mock 不完整，rescue 先修测试基础设施
 
 ---
 
@@ -131,7 +146,13 @@ investigator:
     # 现有的保留，新增：
     - 非首轮：先 review 上轮 git diff，判断方向是否正确
     - 如果上轮 fix 没有让任何测试从 red→green，警告"无进展"
-    - 分类时标注严重度：CRITICAL（编译不过）> HIGH（逻辑错误）> LOW（widget 样式）
+    - 分类时标注严重度和类型：
+      CRITICAL: 编译错误（阻塞一切）
+      HIGH-INFRA: mock 不完整导致 test hang（阻塞行为验证，rescue 优先修）
+      HIGH-LOGIC: 逻辑测试 fail（行为不一致）
+      MEDIUM: widget 测试 fail（渲染/导航差异）
+      LOW: 样式差异
+    - hung 测试 ≠ failed 测试。hung = mock 基础设施问题，不是 v2 行为问题
 ```
 
 ### 收敛保护
@@ -144,9 +165,19 @@ rescue:
     输出卡点分析 + 剩余问题清单
   round_budget: |
     每轮最多改 5 个文件
-    每轮必须让至少 1 个测试从 red/skip → green，否则视为无进展
+    每轮必须让至少 1 个测试从 red/skip/hung → green，否则视为无进展
+  priority_order: |
+    Phase 1: 修 CRITICAL（编译错误）
+    Phase 2: 修 HIGH-INFRA（mock 基础设施，让 hung 测试能跑）
+    Phase 3: 修 HIGH-LOGIC（逻辑测试，让 red 变 green）
+    Phase 4: 修 MEDIUM（widget 测试，让 red 变 green）
+    每个 phase 完成后才进入下一个
 ```
 
+**为什么 HIGH-INFRA 优先于 HIGH-LOGIC？**
+
+hung 的测试 = 行为没被验证 = 重写目标的盲区。如果跳过 mock 修复直接改逻辑，可能 20 个 widget 行为从来没被检验过。先让所有测试能跑（pass 或 fail），再让它们变绿。
+
 ### 完整 rescue workflow
 
 ```yaml