cc-devflow 1.0.2 → 2.4.0

package/.claude/scripts/verify-gate.sh

@@ -0,0 +1,269 @@
+#!/bin/bash
+# ============================================================
+# Verification Gate Script
+# Enforces "Evidence before assertions" principle
+# ============================================================
+
+set -e
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+# Colors
+RED='\033[0;31m'
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+NC='\033[0m'
+
+usage() {
+    echo "Usage: $0 --type TYPE [OPTIONS]"
+    echo ""
+    echo "Types:"
+    echo "  prd         Verify PRD completion"
+    echo "  epic        Verify EPIC completion"
+    echo "  dev         Verify development completion"
+    echo "  qa          Verify QA completion"
+    echo "  release     Verify release readiness"
+    echo ""
+    echo "Options:"
+    echo "  --req-dir DIR    Requirement directory"
+    echo "  --verbose        Show detailed output"
+    echo "  --json           Output as JSON"
+}
+
+verify_prd() {
+    local req_dir="$1"
+    local errors=0
+
+    echo -e "${YELLOW}Verifying PRD completion...${NC}"
+
+    # Check PRD.md exists
+    if [[ ! -f "${req_dir}/PRD.md" ]]; then
+        echo -e "${RED}ERROR: PRD.md not found${NC}"
+        ((errors++))
+    else
+        echo -e "${GREEN}✓ PRD.md exists${NC}"
+    fi
+
+    # Check BRAINSTORM.md exists
+    if [[ ! -f "${req_dir}/BRAINSTORM.md" ]]; then
+        echo -e "${RED}ERROR: BRAINSTORM.md not found${NC}"
+        ((errors++))
+    else
+        echo -e "${GREEN}✓ BRAINSTORM.md exists${NC}"
+    fi
+
+    # Check for TODO/FIXME in PRD
+    if grep -q "TODO\|FIXME\|{{" "${req_dir}/PRD.md" 2>/dev/null; then
+        echo -e "${RED}ERROR: PRD.md contains TODO/FIXME/placeholders${NC}"
+        ((errors++))
+    else
+        echo -e "${GREEN}✓ PRD.md has no placeholders${NC}"
+    fi
+
+    return $errors
+}
+
+verify_epic() {
+    local req_dir="$1"
+    local errors=0
+
+    echo -e "${YELLOW}Verifying EPIC completion...${NC}"
+
+    # Check EPIC.md exists
+    if [[ ! -f "${req_dir}/EPIC.md" ]]; then
+        echo -e "${RED}ERROR: EPIC.md not found${NC}"
+        ((errors++))
+    else
+        echo -e "${GREEN}✓ EPIC.md exists${NC}"
+    fi
+
+    # Check TASKS.md exists
+    if [[ ! -f "${req_dir}/TASKS.md" ]]; then
+        echo -e "${RED}ERROR: TASKS.md not found${NC}"
+        ((errors++))
+    else
+        echo -e "${GREEN}✓ TASKS.md exists${NC}"
+    fi
+
+    return $errors
+}
+
+verify_dev() {
+    local req_dir="$1"
+    local errors=0
+
+    echo -e "${YELLOW}Verifying development completion...${NC}"
+
+    # Run tests
+    echo "Running tests..."
+    if npm test 2>/dev/null; then
+        echo -e "${GREEN}✓ Tests pass${NC}"
+    else
+        echo -e "${RED}ERROR: Tests failed${NC}"
+        ((errors++))
+    fi
+
+    # Run build
+    echo "Running build..."
+    if npm run build 2>/dev/null; then
+        echo -e "${GREEN}✓ Build succeeds${NC}"
+    else
+        echo -e "${RED}ERROR: Build failed${NC}"
+        ((errors++))
+    fi
+
+    # Run lint
+    echo "Running lint..."
+    if npm run lint 2>/dev/null; then
+        echo -e "${GREEN}✓ Lint passes${NC}"
+    else
+        echo -e "${YELLOW}WARNING: Lint issues found${NC}"
+    fi
+
+    return $errors
+}
+
+verify_qa() {
+    local req_dir="$1"
+    local errors=0
+
+    echo -e "${YELLOW}Verifying QA completion...${NC}"
+
+    # Check TEST_REPORT.md exists
+    if [[ ! -f "${req_dir}/TEST_REPORT.md" ]]; then
+        echo -e "${RED}ERROR: TEST_REPORT.md not found${NC}"
+        ((errors++))
+    else
+        echo -e "${GREEN}✓ TEST_REPORT.md exists${NC}"
+    fi
+
+    # Check SECURITY_REPORT.md exists
+    if [[ ! -f "${req_dir}/SECURITY_REPORT.md" ]]; then
+        echo -e "${RED}ERROR: SECURITY_REPORT.md not found${NC}"
+        ((errors++))
+    else
+        echo -e "${GREEN}✓ SECURITY_REPORT.md exists${NC}"
+    fi
+
+    return $errors
+}
+
+verify_release() {
+    local req_dir="$1"
+    local errors=0
+
+    echo -e "${YELLOW}Verifying release readiness...${NC}"
+
+    # Check RELEASE_PLAN.md exists
+    if [[ ! -f "${req_dir}/RELEASE_PLAN.md" ]]; then
+        echo -e "${RED}ERROR: RELEASE_PLAN.md not found${NC}"
+        ((errors++))
+    else
+        echo -e "${GREEN}✓ RELEASE_PLAN.md exists${NC}"
+    fi
+
+    # Check git status
+    if [[ -n "$(git status --porcelain)" ]]; then
+        echo -e "${RED}ERROR: Uncommitted changes exist${NC}"
+        ((errors++))
+    else
+        echo -e "${GREEN}✓ Git working directory clean${NC}"
+    fi
+
+    # Check PR status if on feature branch
+    local branch=$(git branch --show-current)
+    if [[ "$branch" == feature/* ]] || [[ "$branch" == bugfix/* ]]; then
+        echo "Checking PR status..."
+        if gh pr checks 2>/dev/null; then
+            echo -e "${GREEN}✓ PR checks pass${NC}"
+        else
+            echo -e "${YELLOW}WARNING: PR checks not available or failing${NC}"
+        fi
+    fi
+
+    return $errors
+}
+
+# Main
+TYPE=""
+REQ_DIR=""
+VERBOSE=false
+JSON=false
+
+while [[ $# -gt 0 ]]; do
+    case $1 in
+        --type)
+            TYPE="$2"
+            shift 2
+            ;;
+        --req-dir)
+            REQ_DIR="$2"
+            shift 2
+            ;;
+        --verbose)
+            VERBOSE=true
+            shift
+            ;;
+        --json)
+            JSON=true
+            shift
+            ;;
+        --help|-h)
+            usage
+            exit 0
+            ;;
+        *)
+            echo "Unknown option: $1"
+            usage
+            exit 1
+            ;;
+    esac
+done
+
+if [[ -z "$TYPE" ]]; then
+    echo "Error: --type is required"
+    usage
+    exit 1
+fi
+
+# Default req_dir to current directory
+REQ_DIR="${REQ_DIR:-.}"
+
+case "$TYPE" in
+    prd)
+        verify_prd "$REQ_DIR"
+        ;;
+    epic)
+        verify_epic "$REQ_DIR"
+        ;;
+    dev)
+        verify_dev "$REQ_DIR"
+        ;;
+    qa)
+        verify_qa "$REQ_DIR"
+        ;;
+    release)
+        verify_release "$REQ_DIR"
+        ;;
+    *)
+        echo "Unknown type: $TYPE"
+        usage
+        exit 1
+        ;;
+esac
+
+exit_code=$?
+
+if [[ $exit_code -eq 0 ]]; then
+    echo ""
+    echo -e "${GREEN}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━${NC}"
+    echo -e "${GREEN}VERIFICATION PASSED${NC}"
+    echo -e "${GREEN}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━${NC}"
+else
+    echo ""
+    echo -e "${RED}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━${NC}"
+    echo -e "${RED}VERIFICATION FAILED - ${exit_code} error(s)${NC}"
+    echo -e "${RED}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━${NC}"
+fi
+
+exit $exit_code

package/.claude/skills/cc-devflow-orchestrator/SKILL.md

@@ -16,38 +16,51 @@ Guide users to the correct agent/command WITHOUT duplicating their detailed stan
 /core-roadmap → ROADMAP.md + BACKLOG.md (产品路线图)
 /core-architecture → ARCHITECTURE.md (系统架构设计)
 /core-guidelines → frontend-guidelines.md / backend-guidelines.md (项目规范)
-/core-style → STYLE.md (设计风格指南) ⭐ 新增
+/core-style → STYLE.md (设计风格指南) 
 ```
 
 ### 📦 需求级工作流（Requirement-Level, 每个需求执行一次）
 
 ```
-/flow-init → research.md + tasks.json (研究初始化)
+/flow-init → research.md + tasks.json + BRAINSTORM.md (研究初始化 + 头脑风暴)
      ↓
-/flow-clarify → clarifications/*.md (11 维度歧义扫描, 可选) ⭐ 新增
+/flow-clarify → clarifications/*.md (11 维度歧义扫描, 可选)
      ↓
-/flow-prd → PRD.md (invoke prd-writer agent)
+/flow-prd → PRD.md (invoke prd-writer agent, 需 BRAINSTORM.md 对齐)
      ↓
-/flow-checklist → checklists/*.md (需求质量检查, 可选) ⭐ 新增
+/flow-checklist → checklists/*.md (需求质量检查, 可选)
      ↓
 /flow-tech → TECH_DESIGN.md + data-model + contracts (invoke tech-architect agent)
      ↓
-/flow-ui → UI_PROTOTYPE.html (invoke ui-designer agent, 可选, 引用 STYLE.md) ⭐ 变更
+/flow-ui → UI_PROTOTYPE.html (invoke ui-designer agent, 可选, 引用 STYLE.md)
      ↓
-/flow-epic → EPIC.md + TASKS.md (invoke planner agent with PRD+TECH+UI)
+/flow-epic → EPIC.md + TASKS.md (invoke planner, bite-sized tasks)
      ↓
-/flow-dev → TASKS.md execution (TDD order enforced, 引用 STYLE.md) ⭐ 变更
+/flow-dev → TASKS.md execution (TDD + Autonomous mode default)
+     ↓
+/flow-review → SPEC_REVIEW.md + CODE_QUALITY_REVIEW.md (Two-Stage Review) 
      ↓
 /flow-qa → QA reports (invoke qa-tester + security-reviewer agents)
      ↓
-/flow-release → PR creation + deployment
+/flow-release → PR creation + deployment (分支完成决策)
      ↓
 /flow-verify → consistency check (invoke consistency-checker agent, 任意阶段可调用)
 ```
 
+### 🐛 Bug 修复工作流
+
+```
+/flow-fix "BUG-123|描述" → 系统化调试 (4阶段: Root Cause → Pattern → Hypothesis → TDD Fix)
+```
+
 **说明**:
 - 项目级命令建立全局标准（SSOT），需求级命令引用这些标准
+- `/flow-init` 包含 Brainstorming 阶段，生成 BRAINSTORM.md 作为需求「北极星」
+- `/flow-prd` 需要 BRAINSTORM.md 对齐检查
 - `/flow-clarify` 在 PRD 前可选执行，消除 research.md 中的歧义
+- `/flow-epic` 使用 bite-sized tasks 原则 (2-5分钟/任务)
+- `/flow-dev` 默认 Autonomous 模式（自动重试），使用 `--manual` 退出到 Manual 模式 
+- `/flow-review` 是新增的两阶段审查 (Spec Compliance → Code Quality)
 - `/flow-ui` 和 `/flow-dev` 自动加载 `devflow/STYLE.md`（如存在）
 - 项目级命令可按需执行，无严格顺序要求
 
@@ -104,26 +117,43 @@ Guide users to the correct agent/command WITHOUT duplicating their detailed stan
 - **DON'T**: Duplicate QA standards (qa-tester agent has ~300 lines)
 - **Link**: See [.claude/agents/qa-tester.md](.claude/agents/qa-tester.md) for QA details
 
+### When User Asks About Code Review (v2.1.0 新增)
+- **DO**: Recommend `/flow-review` command → invokes spec-reviewer + code-quality-reviewer agents
+- **DON'T**: Duplicate review standards (Two-Stage Review)
+- **Link**: See [.claude/commands/flow-review.md](.claude/commands/flow-review.md) for details
+- **Features**: Stage 1 (Spec Compliance) → Stage 2 (Code Quality), 不信任实现者报告
+
+### When User Asks About Bug Fix (v2.1.0 新增)
+- **DO**: Recommend `/flow-fix` command → 4-phase systematic debugging
+- **DON'T**: Guess and fix without investigation
+- **Link**: See [.claude/commands/flow-fix.md](.claude/commands/flow-fix.md) for details
+- **Features**: Root Cause → Pattern → Hypothesis → TDD Fix, Iron Law enforcement
+
 ## Phase Gates (Quick Reference Only)
 
 ### Entry Gates
+- **flow-init Entry**: Git 工作区干净, main 分支
 - **flow-clarify Entry**: research.md 存在, phase0_complete == true
-- **flow-prd Entry**: research.md 无 TODO placeholder, phase0_complete == true (clarify 可选)
+- **flow-prd Entry**: BRAINSTORM.md 存在, research.md 无 TODO placeholder 
 - **flow-checklist Entry**: PRD.md 必须完成 (prd_complete == true)
 - **flow-tech Entry**: PRD.md 必须完成
 - **flow-ui Entry**: PRD.md 必须完成（可与 tech 并行）
 - **flow-epic Entry**: PRD 完成，tech/ui 推荐但可选，Checklist Gate (如存在 checklists/)
 - **flow-dev Entry**: EPIC.md + TASKS.md 存在
-- **flow-qa Entry**: development_complete == true
+- **flow-review Entry**: development_complete == true 
+- **flow-qa Entry**: review_complete == true (或 development_complete)
+- **flow-release Entry**: qa_complete == true
 
 ### Exit Gates
-- **flow-init Exit**: research.md 5-level quality check
+- **flow-init Exit**: research.md 5-level quality check, BRAINSTORM.md 完整
 - **flow-clarify Exit**: clarification report 完整, orchestration_status.clarify_complete == true
-- **flow-prd Exit**: PRD.md 无 placeholder, Constitution 合规
+- **flow-prd Exit**: PRD.md 无 placeholder, Constitution 合规, BRAINSTORM 对齐
 - **flow-tech Exit**: TECH_DESIGN.md + data-model + contracts 完整
-- **flow-epic Exit**: TASKS.md TDD 顺序正确, Phase -1 Gates 通过
-- **flow-dev Exit**: 所有 TASKS 完成, 测试通过
+- **flow-epic Exit**: TASKS.md TDD 顺序正确, bite-sized tasks, Phase -1 Gates 通过
+- **flow-dev Exit**: 所有 TASKS 完成, TDD Checkpoint 通过, 测试通过
+- **flow-review Exit**: SPEC_REVIEW.md + CODE_QUALITY_REVIEW.md 均 PASS 
 - **flow-qa Exit**: 无 high-severity 漏洞
+- **flow-release Exit**: PR 创建成功, 分支决策完成 
 
 **For Details**: See [orchestration_status.json](devflow/requirements/REQ-XXX/orchestration_status.json) and [EXECUTION_LOG.md](devflow/requirements/REQ-XXX/EXECUTION_LOG.md)
 
@@ -135,6 +165,7 @@ Read `orchestration_status.json` to determine current phase:
 status: "initialized"
   → Recommend: /flow-clarify (optional, clarify ambiguities)
   → Alternative: /flow-prd (skip clarification, generate PRD directly)
+  → Note: BRAINSTORM.md 已在 /flow-init 生成 
 
 status: "clarify_complete" OR "clarify_skipped"
   → Recommend: /flow-prd (generate PRD)
@@ -148,9 +179,14 @@ status: "tech_design_complete"
   → Else: /flow-epic (generate EPIC and TASKS)
 
 status: "epic_complete"
-  → Recommend: /flow-dev (start TDD development)
+  → Recommend: /flow-dev (TDD development, Autonomous mode default)
+  → Alternative: /flow-dev --manual (Manual mode for complex requirements) 
 
 status: "development_complete"
+  → Recommend: /flow-review (Two-Stage Code Review) 
+  → Alternative: /flow-qa (skip review, go directly to QA)
+
+status: "review_complete"
   → Recommend: /flow-qa (quality assurance and security review)
 
 status: "qa_complete"
@@ -181,11 +217,13 @@ status: "released"
 ### Constitution violation?
 - **Real-time check**: constitution-guardian guardrail (PreToolUse hook)
 - **Batch validation**: Run `.claude/scripts/validate-constitution.sh`
-- **Reference**: See `.claude/rules/project-constitution.md` v2.0.0
+- **Reference**: See `.claude/rules/project-constitution.md` 
+- **Rationalization Library**: See `.claude/rules/rationalization-library.md` 
 
 ### TDD order violated?
 - **Real-time check**: devflow-tdd-enforcer guardrail (PreToolUse hook)
 - **Manual check**: See TASKS.md, tests MUST be marked [x] before implementation
+- **TDD Skill**: See `.claude/skills/flow-tdd/SKILL.md` 
 
 ## Auxiliary Commands
 
@@ -200,16 +238,28 @@ status: "released"
 - `/flow-verify "REQ-123"` - Comprehensive consistency verification
 
 ### Bug Fix
-- `/flow-fix "BUG-123|登录超时"` - One-shot BUG fix workflow
+- `/flow-fix "BUG-123|登录超时"` - 系统化 BUG 修复 (4阶段调试法)
 - `/problem-analyzer "<issue>"` - Problem diagnosis
 
 ### Code Review
+- `/flow-review "REQ-123"` - Two-Stage Code Review (Spec → Quality)
 - `/code-review-high "<diff>"` - High-rigor code review
 
 ## Integration with Other Skills
 
-- **devflow-tdd-enforcer**: Enforces TDD order in TASKS.md (Guardrail, blocks)
-- **constitution-guardian**: Enforces Constitution compliance (Guardrail, blocks)
+### Guardrails (实时阻断)
+- **devflow-tdd-enforcer**: Enforces TDD order in TASKS.md
+- **constitution-guardian**: Enforces Constitution compliance
+
+### Workflow Skills
+- **flow-brainstorming**: 需求头脑风暴，生成 BRAINSTORM.md
+- **flow-tdd**: TDD Iron Law 执行
+- **flow-debugging**: 4阶段系统化调试
+- **flow-receiving-review**: 处理代码审查反馈
+- **flow-finishing-branch**: 分支完成决策
+- **verification-before-completion**: 验证闸门
+
+### Reference Skills
 - **devflow-file-standards**: File naming and directory structure reference
 - **devflow-constitution-quick-ref**: Constitution quick reference
 

package/.claude/skills/file-header-guardian/SKILL.md

@@ -0,0 +1,56 @@
+---
+name: file-header-guardian
+description: 文件头三行契约注释。触发：create file、新建文件、编写代码。
+---
+
+# 文件头注释守护者
+
+触发：创建代码文件，修改文件后检查头注释准确性。
+
+## 模板
+
+**TS/JS:**
+```typescript
+/**
+ * @input  依赖什么
+ * @output 提供什么
+ * @pos    系统地位
+ * ⚠️ 修改后同步：文件头 + 目录 CLAUDE.md
+ */
+```
+
+**Python:**
+```python
+"""
+@input  依赖什么
+@output 提供什么
+@pos    系统地位
+⚠️ 修改后同步：文件头 + 目录 CLAUDE.md
+"""
+```
+
+**Shell:**
+```bash
+# @input  依赖什么
+# @output 提供什么
+# @pos    系统地位
+# ⚠️ 修改后同步：文件头 + 目录 CLAUDE.md
+```
+
+## 示例
+
+```typescript
+/**
+ * @input  prisma.service 数据库连接
+ * @output UserService: create/update/delete
+ * @pos    用户模块核心
+ * ⚠️ 修改后同步：文件头 + 目录 CLAUDE.md
+ */
+export class UserService { }
+```
+
+## 豁免
+
+配置文件（json/yaml）、样式（css）、生成文件（*.d.ts）、node_modules。
+
+协作：本 skill 管文件，fractal-docs-generator 管目录。

package/.claude/skills/flow-attention-refresh/SKILL.md

@@ -0,0 +1,170 @@
+---
+name: flow-attention-refresh
+description: "注意力刷新协议。在关键时刻强制读取核心文档，防止目标遗忘。被 flow-dev, flow-ralph 等命令引用。"
+---
+
+# Flow Attention Refresh - 注意力刷新协议
+
+## Overview
+
+解决 "Lost in the Middle" 效应：当上下文过长时，原始目标被"推出"注意力窗口。
+
+```
+问题:
+上下文开始: [原始目标 - 距离很远，被"遗忘"]
+     ↓
+... 50+ 工具调用 ...
+     ↓
+上下文末尾: [最近的工具输出 - 在注意力窗口]
+
+解决:
+在关键时刻强制读取目标文档 → 将目标"拉回"注意力窗口
+```
+
+## The Iron Law
+
+```
+BEFORE MAJOR DECISION → READ GOAL FILES → THEN ACT
+```
+
+## Refresh Protocols
+
+### Protocol 1: Flow Entry (每个 flow-* 命令开始)
+
+**触发点**: 任何 `/flow-*` 命令的 Entry Gate
+
+**动作**:
+```yaml
+read:
+  - BRAINSTORM.md "成功标准" 章节
+  - TASKS.md 当前阶段任务列表
+focus: 本次命令要达成什么？
+```
+
+**实现**: 已由现有 Brainstorm Alignment Check 覆盖
+
+---
+
+### Protocol 2: Task Start (flow-dev 每个任务开始)
+
+**触发点**: `flow-dev` TDD 循环中，每个任务执行前
+
+**动作**:
+```yaml
+read:
+  - TASKS.md 当前任务 T### 段落 + DoD
+  - quickstart.md 相关命令
+focus: 这个任务的验收标准是什么？
+```
+
+**实现位置**: flow-dev.md 阶段 3 TDD 循环
+
+**代码片段**:
+```markdown
+**注意力刷新** (Protocol 2):
+→ Read: TASKS.md 当前任务 T### 段落
+→ Focus: 任务目标和 DoD
+→ Then: 开始执行任务
+```
+
+---
+
+### Protocol 3: Ralph Iteration (Ralph 循环每次迭代)
+
+**触发点**: `/flow-ralph` 每次迭代开始
+
+**动作**:
+```yaml
+read:
+  - TASKS.md 第一个未完成任务
+  - ERROR_LOG.md 最近 5 条记录
+focus: 下一步行动 + 避免重复错误
+```
+
+**实现位置**: flow-ralph.md Ralph Loop 开始处
+
+**代码片段**:
+```markdown
+**注意力刷新** (Protocol 3):
+→ Read: TASKS.md 找到第一个 `- [ ]` 任务
+→ Read: ERROR_LOG.md 最近 5 条（如存在）
+→ Focus: 下一步行动是什么？有什么错误需要避免？
+→ Then: 执行任务
+```
+
+---
+
+### Protocol 4: Error Recovery (遇到错误后)
+
+**触发点**: 测试失败、构建错误、运行时异常后
+
+**动作**:
+```yaml
+read:
+  - ERROR_LOG.md 当前错误记录
+  - TASKS.md 当前任务定义
+focus: 错误根因 + 解决方案
+```
+
+**实现位置**: flow-tdd skill 错误处理部分
+
+**代码片段**:
+```markdown
+**错误恢复刷新** (Protocol 4):
+→ Record: 将错误追加到 ERROR_LOG.md
+→ Read: ERROR_LOG.md 当前错误 + 历史相似错误
+→ Read: TASKS.md 当前任务
+→ Focus: 根因分析，避免盲目修复
+→ Then: 有针对性地修复
+```
+
+---
+
+## Integration Map
+
+```
+现有机制:
+┌──────────────────────────────────────┐
+│ flow-brainstorming skill             │
+│ ↓ 触发于 /flow-init                  │
+│ ↓ 输出 BRAINSTORM.md                 │
+└──────────────────────────────────────┘
+              ↓
+┌──────────────────────────────────────┐
+│ Brainstorm Alignment Check           │
+│ ↓ 在每个 flow-* Entry Gate           │
+│ ↓ 读取 BRAINSTORM.md 验证对齐        │
+│ ↓ = Protocol 1                       │
+└──────────────────────────────────────┘
+
+新增机制:
+┌──────────────────────────────────────┐
+│ flow-attention-refresh skill (本文件) │
+│ ↓ 被 flow-dev, flow-ralph 引用       │
+│ ↓ 在更细粒度的时刻刷新注意力          │
+│ ↓ Protocol 2, 3, 4                   │
+└──────────────────────────────────────┘
+```
+
+## Rationalization Prevention
+
+| Excuse | Reality |
+|--------|---------|
+| "我记得目标是什么" | 上下文 50+ 工具调用后，你真的不记得了 |
+| "读文件浪费时间" | 读文件 < 1秒，返工 > 10分钟 |
+| "这个任务很简单" | 简单任务也会偏离目标 |
+| "刚刚读过了" | 每次迭代都要读，这是协议 |
+
+## Red Flags - STOP
+
+如果你发现自己：
+- 开始任务前没有读取 TASKS.md
+- 遇到错误后没有先记录 ERROR_LOG.md
+- Ralph 循环中忘记读取上次错误
+- 做决策时没有参考 BRAINSTORM.md
+
+**STOP。执行对应的 Protocol。**
+
+---
+
+**[PROTOCOL]**: 变更时更新此头部，然后检查 CLAUDE.md