diff --git a/.claude/skills/command-guide/index/all-commands.json b/.claude/skills/command-guide/index/all-commands.json
index ab8a82bc..449bdd22 100644
--- a/.claude/skills/command-guide/index/all-commands.json
+++ b/.claude/skills/command-guide/index/all-commands.json
@@ -112,7 +112,7 @@
   {
     "name": "tech-research",
     "command": "/memory:tech-research",
-    "description": "3-phase orchestrator: extract tech stack from session/name → delegate to agent for Exa research and module generation → generate SKILL.md index (skips phase 2 if exists)",
+    "description": "3-phase orchestrator: extract tech stack → Exa research → generate path-conditional rules (auto-loaded by Claude Code)",
     "arguments": "[session-id | tech-stack-name] [--regenerate] [--tool <gemini|qwen>]",
     "category": "memory",
     "subcategory": null,
diff --git a/.claude/skills/command-guide/index/by-category.json b/.claude/skills/command-guide/index/by-category.json
index 5f678d3a..8c9202d2 100644
--- a/.claude/skills/command-guide/index/by-category.json
+++ b/.claude/skills/command-guide/index/by-category.json
@@ -133,7 +133,7 @@
       {
         "name": "tech-research",
         "command": "/memory:tech-research",
-        "description": "3-phase orchestrator: extract tech stack from session/name → delegate to agent for Exa research and module generation → generate SKILL.md index (skips phase 2 if exists)",
+        "description": "3-phase orchestrator: extract tech stack → Exa research → generate path-conditional rules (auto-loaded by Claude Code)",
         "arguments": "[session-id | tech-stack-name] [--regenerate] [--tool <gemini|qwen>]",
         "category": "memory",
         "subcategory": null,
diff --git a/.claude/skills/command-guide/index/by-use-case.json b/.claude/skills/command-guide/index/by-use-case.json
index 2032d50a..808c5905 100644
--- a/.claude/skills/command-guide/index/by-use-case.json
+++ b/.claude/skills/command-guide/index/by-use-case.json
@@ -36,7 +36,7 @@
     {
       "name": "tech-research",
       "command": "/memory:tech-research",
-      "description": "3-phase orchestrator: extract tech stack from session/name → delegate to agent for Exa research and module generation → generate SKILL.md index (skips phase 2 if exists)",
+      "description": "3-phase orchestrator: extract tech stack → Exa research → generate path-conditional rules (auto-loaded by Claude Code)",
       "arguments": "[session-id | tech-stack-name] [--regenerate] [--tool <gemini|qwen>]",
       "category": "memory",
       "subcategory": null,
diff --git a/.claude/skills/command-guide/reference/agents/action-planning-agent.md b/.claude/skills/command-guide/reference/agents/action-planning-agent.md
index 0fd9db47..ad38cd16 100644
--- a/.claude/skills/command-guide/reference/agents/action-planning-agent.md
+++ b/.claude/skills/command-guide/reference/agents/action-planning-agent.md
@@ -409,14 +409,14 @@ Generate individual `.task/IMPL-*.json` files with the following structure:
   // Pattern: Gemini CLI deep analysis
   {
     "step": "gemini_analyze_[aspect]",
-    "command": "bash(cd [path] && gemini -p 'PURPOSE: [goal]\\nTASK: [tasks]\\nMODE: analysis\\nCONTEXT: @[paths]\\nEXPECTED: [output]\\nRULES: $(cat [template]) | [constraints] | analysis=READ-ONLY')",
+    "command": "ccw cli exec 'PURPOSE: [goal]\\nTASK: [tasks]\\nMODE: analysis\\nCONTEXT: @[paths]\\nEXPECTED: [output]\\nRULES: $(cat [template]) | [constraints] | analysis=READ-ONLY' --tool gemini --cd [path]",
     "output_to": "analysis_result"
   },
 
   // Pattern: Qwen CLI analysis (fallback/alternative)
   {
     "step": "qwen_analyze_[aspect]",
-    "command": "bash(cd [path] && qwen -p '[similar to gemini pattern]')",
+    "command": "ccw cli exec '[similar to gemini pattern]' --tool qwen --cd [path]",
     "output_to": "analysis_result"
   },
 
@@ -457,7 +457,7 @@ The examples above demonstrate **patterns**, not fixed requirements. Agent MUST:
 4. **Command Composition Patterns**:
    - **Single command**: `bash([simple_search])`
    - **Multiple commands**: `["bash([cmd1])", "bash([cmd2])"]`
-   - **CLI analysis**: `bash(cd [path] && gemini -p '[prompt]')`
+   - **CLI analysis**: `ccw cli exec '[prompt]' --tool gemini --cd [path]`
    - **MCP integration**: `mcp__[tool]__[function]([params])`
 
 **Key Principle**: Examples show **structure patterns**, not specific implementations. Agent must create task-appropriate steps dynamically.
@@ -481,9 +481,9 @@ The `implementation_approach` supports **two execution modes** based on the pres
    - **Use for**: Large-scale features, complex refactoring, or when user explicitly requests CLI tool usage
    - **Required fields**: Same as default mode **PLUS** `command`
    - **Command patterns**:
-     - `bash(codex -C [path] --full-auto exec '[prompt]' --skip-git-repo-check -s danger-full-access)`
-     - `bash(codex --full-auto exec '[task]' resume --last --skip-git-repo-check -s danger-full-access)` (multi-step)
-     - `bash(cd [path] && gemini -p '[prompt]' --approval-mode yolo)` (write mode)
+     - `ccw cli exec '[prompt]' --tool codex --mode auto --cd [path]`
+     - `ccw cli exec '[task]' --tool codex --mode auto` (multi-step with context)
+     - `ccw cli exec '[prompt]' --tool gemini --mode write --cd [path]` (write mode)
 
 **Semantic CLI Tool Selection**:
 
@@ -500,12 +500,12 @@ Agent determines CLI tool usage per-step based on user semantics and task nature
 **Task-Based Selection** (when no explicit user preference):
 - **Implementation/coding**: Codex preferred for autonomous development
 - **Analysis/exploration**: Gemini preferred for large context analysis
-- **Documentation**: Gemini/Qwen with write mode (`--approval-mode yolo`)
+- **Documentation**: Gemini/Qwen with write mode (`--mode write`)
 - **Testing**: Depends on complexity - simple=agent, complex=Codex
 
 **Default Behavior**: Agent always executes the workflow. CLI commands are embedded in `implementation_approach` steps:
 - Agent orchestrates task execution
-- When step has `command` field, agent executes it via Bash
+- When step has `command` field, agent executes it via CCW CLI
 - When step has no `command` field, agent implements directly
 - This maintains agent control while leveraging CLI tool power
 
@@ -559,7 +559,7 @@ Agent determines CLI tool usage per-step based on user semantics and task nature
     "step": 3,
     "title": "Execute implementation using CLI tool",
     "description": "Use Codex/Gemini for complex autonomous execution",
-    "command": "bash(codex -C [path] --full-auto exec '[prompt]' --skip-git-repo-check -s danger-full-access)",
+    "command": "ccw cli exec '[prompt]' --tool codex --mode auto --cd [path]",
     "modification_points": ["[Same as default mode]"],
     "logic_flow": ["[Same as default mode]"],
     "depends_on": [1, 2],
diff --git a/.claude/skills/command-guide/reference/agents/cli-execution-agent.md b/.claude/skills/command-guide/reference/agents/cli-execution-agent.md
index f661a8cc..ba50087c 100644
--- a/.claude/skills/command-guide/reference/agents/cli-execution-agent.md
+++ b/.claude/skills/command-guide/reference/agents/cli-execution-agent.md
@@ -100,7 +100,7 @@ CONTEXT: @**/*
 # Specific patterns
 CONTEXT: @CLAUDE.md @src/**/* @*.ts
 
-# Cross-directory (requires --include-directories)
+# Cross-directory (requires --includeDirs)
 CONTEXT: @**/* @../shared/**/* @../types/**/*
 ```
 
@@ -144,43 +144,40 @@ discuss → multi (gemini + codex parallel)
 - Codex: `gpt-5` (default), `gpt5-codex` (large context)
 - **Position**: `-m` after prompt, before flags
 
-### Command Templates
+### Command Templates (CCW Unified CLI)
 
 **Gemini/Qwen (Analysis)**:
 ```bash
-cd {dir} && gemini -p "
+ccw cli exec "
 PURPOSE: {goal}
 TASK: {task}
 MODE: analysis
 CONTEXT: @**/*
 EXPECTED: {output}
 RULES: $(cat ~/.claude/workflows/cli-templates/prompts/analysis/pattern.txt)
-" -m gemini-2.5-pro
+" --tool gemini --cd {dir}
 
-# Qwen fallback: Replace 'gemini' with 'qwen'
+# Qwen fallback: Replace '--tool gemini' with '--tool qwen'
 ```
 
 **Gemini/Qwen (Write)**:
 ```bash
-cd {dir} && gemini -p "..." --approval-mode yolo
+ccw cli exec "..." --tool gemini --mode write --cd {dir}
 ```
 
 **Codex (Auto)**:
 ```bash
-codex -C {dir} --full-auto exec "..." --skip-git-repo-check -s danger-full-access
-
-# Resume: Add 'resume --last' after prompt
-codex --full-auto exec "..." resume --last --skip-git-repo-check -s danger-full-access
+ccw cli exec "..." --tool codex --mode auto --cd {dir}
 ```
 
 **Cross-Directory** (Gemini/Qwen):
 ```bash
-cd src/auth && gemini -p "CONTEXT: @**/* @../shared/**/*" --include-directories ../shared
+ccw cli exec "CONTEXT: @**/* @../shared/**/*" --tool gemini --cd src/auth --includeDirs ../shared
 ```
 
 **Directory Scope**:
 - `@` only references current directory + subdirectories
-- External dirs: MUST use `--include-directories` + explicit CONTEXT reference
+- External dirs: MUST use `--includeDirs` + explicit CONTEXT reference
 
 **Timeout**: Simple 20min | Medium 40min | Complex 60min (Codex ×1.5)
 
diff --git a/.claude/skills/command-guide/reference/agents/cli-explore-agent.md b/.claude/skills/command-guide/reference/agents/cli-explore-agent.md
index 1fd7e2c6..7ba68cb1 100644
--- a/.claude/skills/command-guide/reference/agents/cli-explore-agent.md
+++ b/.claude/skills/command-guide/reference/agents/cli-explore-agent.md
@@ -78,14 +78,14 @@ rg "^import .* from " -n | head -30
 ### Gemini Semantic Analysis (deep-scan, dependency-map)
 
 ```bash
-cd {dir} && gemini -p "
+ccw cli exec "
 PURPOSE: {from prompt}
 TASK: {from prompt}
 MODE: analysis
 CONTEXT: @**/*
 EXPECTED: {from prompt}
 RULES: {from prompt, if template specified} | analysis=READ-ONLY
-"
+" --tool gemini --cd {dir}
 ```
 
 **Fallback Chain**: Gemini → Qwen → Codex → Bash-only
diff --git a/.claude/skills/command-guide/reference/agents/cli-lite-planning-agent.md b/.claude/skills/command-guide/reference/agents/cli-lite-planning-agent.md
index 727e1335..19653aec 100644
--- a/.claude/skills/command-guide/reference/agents/cli-lite-planning-agent.md
+++ b/.claude/skills/command-guide/reference/agents/cli-lite-planning-agent.md
@@ -97,7 +97,7 @@ Phase 3: planObject Generation
 ## CLI Command Template
 
 ```bash
-cd {project_root} && {cli_tool} -p "
+ccw cli exec "
 PURPOSE: Generate implementation plan for {complexity} task
 TASK:
 • Analyze: {task_description}
@@ -134,7 +134,7 @@ RULES: $(cat ~/.claude/workflows/cli-templates/prompts/planning/02-breakdown-tas
 - Acceptance must be quantified (counts, method names, metrics)
 - Dependencies use task IDs (T1, T2)
 - analysis=READ-ONLY
-"
+" --tool {cli_tool} --cd {project_root}
 ```
 
 ## Core Functions
diff --git a/.claude/skills/command-guide/reference/agents/cli-planning-agent.md b/.claude/skills/command-guide/reference/agents/cli-planning-agent.md
index 6abf8c3c..ee7dd55d 100644
--- a/.claude/skills/command-guide/reference/agents/cli-planning-agent.md
+++ b/.claude/skills/command-guide/reference/agents/cli-planning-agent.md
@@ -107,7 +107,7 @@ Phase 3: Task JSON Generation
 
 **Template-Based Command Construction with Test Layer Awareness**:
 ```bash
-cd {project_root} && {cli_tool} -p "
+ccw cli exec "
 PURPOSE: Analyze {test_type} test failures and generate fix strategy for iteration {iteration}
 TASK:
 • Review {failed_tests.length} {test_type} test failures: [{test_names}]
@@ -134,7 +134,7 @@ RULES: $(cat ~/.claude/workflows/cli-templates/prompts/{template}) |
 - Consider previous iteration failures
 - Validate fix doesn't introduce new vulnerabilities
 - analysis=READ-ONLY
-" {timeout_flag}
+" --tool {cli_tool} --cd {project_root} --timeout {timeout_value}
 ```
 
 **Layer-Specific Guidance Injection**:
@@ -527,9 +527,9 @@ See: `.process/iteration-{iteration}-cli-output.txt`
 1. **Detect test_type**: "integration" → Apply integration-specific diagnosis
 2. **Execute CLI**:
    ```bash
-   gemini -p "PURPOSE: Analyze integration test failure...
+   ccw cli exec "PURPOSE: Analyze integration test failure...
    TASK: Examine component interactions, data flow, interface contracts...
-   RULES: Analyze full call stack and data flow across components"
+   RULES: Analyze full call stack and data flow across components" --tool gemini
    ```
 3. **Parse Output**: Extract RCA, 修复建议, 验证建议 sections
 4. **Generate Task JSON** (IMPL-fix-1.json):
diff --git a/.claude/skills/command-guide/reference/agents/code-developer.md b/.claude/skills/command-guide/reference/agents/code-developer.md
index e01a0e28..948ab288 100644
--- a/.claude/skills/command-guide/reference/agents/code-developer.md
+++ b/.claude/skills/command-guide/reference/agents/code-developer.md
@@ -34,10 +34,11 @@ You are a code execution specialist focused on implementing high-quality, produc
 - **context-package.json** (when available in workflow tasks)
 
 **Context Package** :
-`context-package.json` provides artifact paths - extract dynamically using `jq`:
+`context-package.json` provides artifact paths - read using `ccw session`:
 ```bash
-# Get role analysis paths from context package
-jq -r '.brainstorm_artifacts.role_analyses[].files[].path' context-package.json
+# Get context package content from session
+ccw session read ${SESSION_ID} --type context
+# Returns parsed JSON with brainstorm_artifacts, focus_paths, etc.
 ```
 
 **Pre-Analysis: Smart Tech Stack Loading**:
@@ -121,9 +122,9 @@ When task JSON contains `flow_control.implementation_approach` array:
    - If `command` field present, execute it; otherwise use agent capabilities
 
 **CLI Command Execution (CLI Execute Mode)**:
-When step contains `command` field with Codex CLI, execute via Bash tool. For Codex resume:
-- First task (`depends_on: []`): `codex -C [path] --full-auto exec "..." --skip-git-repo-check -s danger-full-access`
-- Subsequent tasks (has `depends_on`): Add `resume --last` flag to maintain session context
+When step contains `command` field with Codex CLI, execute via CCW CLI. For Codex resume:
+- First task (`depends_on: []`): `ccw cli exec "..." --tool codex --mode auto --cd [path]`
+- Subsequent tasks (has `depends_on`): Use CCW CLI with resume context to maintain session
 
 **Test-Driven Development**:
 - Write tests first (red → green → refactor)
diff --git a/.claude/skills/command-guide/reference/agents/doc-generator.md b/.claude/skills/command-guide/reference/agents/doc-generator.md
index d867702d..41683af3 100644
--- a/.claude/skills/command-guide/reference/agents/doc-generator.md
+++ b/.claude/skills/command-guide/reference/agents/doc-generator.md
@@ -61,9 +61,9 @@ The agent supports **two execution modes** based on task JSON's `meta.cli_execut
 
    **Step 2** (CLI execution):
    - Agent substitutes [target_folders] into command
-   - Agent executes CLI command via Bash tool:
+   - Agent executes CLI command via CCW:
    ```bash
-   bash(cd src/modules && gemini --approval-mode yolo -p "
+   ccw cli exec "
    PURPOSE: Generate module documentation
    TASK: Create API.md and README.md for each module
    MODE: write
@@ -71,7 +71,7 @@ The agent supports **two execution modes** based on task JSON's `meta.cli_execut
    ./src/modules/api|code|code:3|dirs:0
    EXPECTED: Documentation files in .workflow/docs/my_project/src/modules/
    RULES: $(cat ~/.claude/workflows/cli-templates/prompts/documentation/module-documentation.txt) | Mirror source structure
-   ")
+   " --tool gemini --mode write --cd src/modules
    ```
 
 4. **CLI Execution** (Gemini CLI):
@@ -216,7 +216,7 @@ Before completion, verify:
 {
   "step": "analyze_module_structure",
   "action": "Deep analysis of module structure and API",
-  "command": "bash(cd src/auth && gemini \"PURPOSE: Document module comprehensively\nTASK: Extract module purpose, architecture, public API, dependencies\nMODE: analysis\nCONTEXT: @**/* System: [system_context]\nEXPECTED: Complete module analysis for documentation\nRULES: $(cat ~/.claude/workflows/cli-templates/prompts/documentation/module-documentation.txt)\")",
+  "command": "ccw cli exec \"PURPOSE: Document module comprehensively\nTASK: Extract module purpose, architecture, public API, dependencies\nMODE: analysis\nCONTEXT: @**/* System: [system_context]\nEXPECTED: Complete module analysis for documentation\nRULES: $(cat ~/.claude/workflows/cli-templates/prompts/documentation/module-documentation.txt)\" --tool gemini --cd src/auth",
   "output_to": "module_analysis",
   "on_error": "fail"
 }
diff --git a/.claude/skills/command-guide/reference/commands/memory/docs.md b/.claude/skills/command-guide/reference/commands/memory/docs.md
index 510065d3..d8d91f52 100644
--- a/.claude/skills/command-guide/reference/commands/memory/docs.md
+++ b/.claude/skills/command-guide/reference/commands/memory/docs.md
@@ -191,10 +191,10 @@ Large Projects (single dir >10 docs):
 
 ```bash
 # 1. Get top-level directories from doc-planning-data.json
-bash(cat .workflow/active/WFS-docs-{timestamp}/.process/doc-planning-data.json | jq -r '.top_level_dirs[]')
+ccw session read WFS-docs-{timestamp} --type process --filename doc-planning-data.json --raw | jq -r '.top_level_dirs[]'
 
 # 2. Get mode from workflow-session.json
-bash(cat .workflow/active/WFS-docs-{timestamp}/workflow-session.json | jq -r '.mode // "full"')
+ccw session read WFS-docs-{timestamp} --type session --raw | jq -r '.mode // "full"'
 
 # 3. Check for HTTP API
 bash(grep -r "router\.|@Get\|@Post" src/ 2>/dev/null && echo "API_FOUND" || echo "NO_API")
@@ -223,7 +223,7 @@ bash(grep -r "router\.|@Get\|@Post" src/ 2>/dev/null && echo "API_FOUND" || echo
 
 **Task ID Calculation**:
 ```bash
-group_count=$(jq '.groups.count' .workflow/active/WFS-docs-{timestamp}/.process/doc-planning-data.json)
+group_count=$(ccw session read WFS-docs-{timestamp} --type process --filename doc-planning-data.json --raw | jq '.groups.count')
 readme_id=$((group_count + 1))   # Next ID after groups
 arch_id=$((group_count + 2))
 api_id=$((group_count + 3))
@@ -236,12 +236,12 @@ api_id=$((group_count + 3))
 | Mode | cli_execute | Placement | CLI MODE | Approval Flag | Agent Role |
 |------|-------------|-----------|----------|---------------|------------|
 | **Agent** | false | pre_analysis | analysis | (none) | Generate docs in implementation_approach |
-| **CLI** | true | implementation_approach | write | --approval-mode yolo | Execute CLI commands, validate output |
+| **CLI** | true | implementation_approach | write | --mode write | Execute CLI commands, validate output |
 
 **Command Patterns**:
-- Gemini/Qwen: `cd dir && gemini -p "..."`
-- CLI Mode: `cd dir && gemini --approval-mode yolo -p "..."`
-- Codex: `codex -C dir --full-auto exec "..." --skip-git-repo-check -s danger-full-access`
+- Gemini/Qwen: `ccw cli exec "..." --tool gemini --cd dir`
+- CLI Mode: `ccw cli exec "..." --tool gemini --mode write --cd dir`
+- Codex: `ccw cli exec "..." --tool codex --mode auto --cd dir`
 
 **Generation Process**:
 1. Read configuration values (tool, cli_execute, mode) from workflow-session.json
@@ -286,8 +286,8 @@ api_id=$((group_count + 3))
         "step": "load_precomputed_data",
         "action": "Load Phase 2 analysis and extract group directories",
         "commands": [
-          "bash(cat ${session_dir}/.process/doc-planning-data.json)",
-          "bash(jq '.groups.assignments[] | select(.group_id == \"${group_number}\") | .directories' ${session_dir}/.process/doc-planning-data.json)"
+          "ccw session read ${session_id} --type process --filename doc-planning-data.json",
+          "ccw session read ${session_id} --type process --filename doc-planning-data.json --raw | jq '.groups.assignments[] | select(.group_id == \"${group_number}\") | .directories'"
         ],
         "output_to": "phase2_context",
         "note": "Single JSON file contains all Phase 2 analysis results"
@@ -332,7 +332,7 @@ api_id=$((group_count + 3))
 {
   "step": 2,
   "title": "Batch generate documentation via CLI",
-  "command": "bash(dirs=$(jq -r '.groups.assignments[] | select(.group_id == \"${group_number}\") | .directories[]' ${session_dir}/.process/doc-planning-data.json); for dir in $dirs; do cd \"$dir\" && gemini --approval-mode yolo -p \"PURPOSE: Generate module docs\\nTASK: Create documentation\\nMODE: write\\nCONTEXT: @**/* [phase2_context]\\nEXPECTED: API.md and README.md\\nRULES: Mirror structure\" || echo \"Failed: $dir\"; cd -; done)",
+  "command": "ccw cli exec 'PURPOSE: Generate module docs\\nTASK: Create documentation\\nMODE: write\\nCONTEXT: @**/* [phase2_context]\\nEXPECTED: API.md and README.md\\nRULES: Mirror structure' --tool gemini --mode write --cd ${dirs_from_group}",
   "depends_on": [1],
   "output": "generated_docs"
 }
@@ -602,7 +602,7 @@ api_id=$((group_count + 3))
 | Mode | CLI Placement | CLI MODE | Approval Flag | Agent Role |
 |------|---------------|----------|---------------|------------|
 | **Agent (default)** | pre_analysis | analysis | (none) | Generates documentation content |
-| **CLI (--cli-execute)** | implementation_approach | write | --approval-mode yolo | Executes CLI commands, validates output |
+| **CLI (--cli-execute)** | implementation_approach | write | --mode write | Executes CLI commands, validates output |
 
 **Execution Flow**:
 - **Phase 2**: Unified analysis once, results in `.process/`
diff --git a/.claude/skills/command-guide/reference/commands/memory/load.md b/.claude/skills/command-guide/reference/commands/memory/load.md
index daacae82..be1be51f 100644
--- a/.claude/skills/command-guide/reference/commands/memory/load.md
+++ b/.claude/skills/command-guide/reference/commands/memory/load.md
@@ -5,7 +5,7 @@ argument-hint: "[--tool gemini|qwen] \"task context description\""
 allowed-tools: Task(*), Bash(*)
 examples:
   - /memory:load "在当前前端基础上开发用户认证功能"
-  - /memory:load --tool qwen -p "重构支付模块API"
+  - /memory:load --tool qwen "重构支付模块API"
 ---
 
 # Memory Load Command (/memory:load)
@@ -136,7 +136,7 @@ Task(
 Execute Gemini/Qwen CLI for deep analysis (saves main thread tokens):
 
 \`\`\`bash
-cd . && ${tool} -p "
+ccw cli exec "
 PURPOSE: Extract project core context for task: ${task_description}
 TASK: Analyze project architecture, tech stack, key patterns, relevant files
 MODE: analysis
@@ -147,7 +147,7 @@ RULES:
 - Identify key architecture patterns and technical constraints
 - Extract integration points and development standards
 - Output concise, structured format
-"
+" --tool ${tool}
 \`\`\`
 
 ### Step 4: Generate Core Content Package
@@ -212,7 +212,7 @@ Before returning:
 ### Example 2: Using Qwen Tool
 
 ```bash
-/memory:load --tool qwen -p "重构支付模块API"
+/memory:load --tool qwen "重构支付模块API"
 ```
 
 Agent uses Qwen CLI for analysis, returns same structured package.
diff --git a/.claude/skills/command-guide/reference/commands/memory/tech-research.md b/.claude/skills/command-guide/reference/commands/memory/tech-research.md
index bb3eae8d..43665ca0 100644
--- a/.claude/skills/command-guide/reference/commands/memory/tech-research.md
+++ b/.claude/skills/command-guide/reference/commands/memory/tech-research.md
@@ -1,477 +1,314 @@
 ---
 name: tech-research
-description: 3-phase orchestrator: extract tech stack from session/name → delegate to agent for Exa research and module generation → generate SKILL.md index (skips phase 2 if exists)
+description: "3-phase orchestrator: extract tech stack → Exa research → generate path-conditional rules (auto-loaded by Claude Code)"
 argument-hint: "[session-id | tech-stack-name] [--regenerate] [--tool <gemini|qwen>]"
 allowed-tools: SlashCommand(*), TodoWrite(*), Bash(*), Read(*), Write(*), Task(*)
 ---
 
-# Tech Stack Research SKILL Generator
+# Tech Stack Rules Generator
 
 ## Overview
 
-**Pure Orchestrator with Agent Delegation**: Prepares context paths and delegates ALL work to agent. Agent produces files directly.
+**Purpose**: Generate multi-layered, path-conditional rules that Claude Code automatically loads based on file context.
 
-**Auto-Continue Workflow**: Runs fully autonomously once triggered. Each phase completes and automatically triggers the next phase.
+**Key Difference from SKILL Memory**:
+- **SKILL**: Manual loading via `Skill(command: "tech-name")`
+- **Rules**: Automatic loading when working with matching file paths
 
-**Execution Paths**:
-- **Full Path**: All 3 phases (no existing SKILL OR `--regenerate` specified)
-- **Skip Path**: Phase 1 → Phase 3 (existing SKILL found AND no `--regenerate` flag)
-- **Phase 3 Always Executes**: SKILL index is always generated or updated
+**Output Structure**:
+```
+.claude/rules/tech/{tech-stack}/
+├── core.md           # paths: **/*.{ext} - Core principles
+├── patterns.md       # paths: src/**/*.{ext} - Implementation patterns
+├── testing.md        # paths: **/*.{test,spec}.{ext} - Testing rules
+├── config.md         # paths: *.config.* - Configuration rules
+├── api.md            # paths: **/api/**/* - API rules (backend only)
+├── components.md     # paths: **/components/**/* - Component rules (frontend only)
+└── metadata.json     # Generation metadata
+```
 
-**Agent Responsibility**:
-- Agent does ALL the work: context reading, Exa research, content synthesis, file writing
-- Orchestrator only provides context paths and waits for completion
+**Templates Location**: `~/.claude/workflows/cli-templates/prompts/rules/`
+
+---
 
 ## Core Rules
 
-1. **Start Immediately**: First action is TodoWrite initialization, second action is Phase 1 execution
-2. **Context Path Delegation**: Pass session directory or tech stack name to agent, let agent do discovery
-3. **Agent Produces Files**: Agent directly writes all module files, orchestrator does NOT parse agent output
-4. **Auto-Continue**: After completing each phase, update TodoWrite and immediately execute next phase
-5. **No User Prompts**: Never ask user questions or wait for input between phases
-6. **Track Progress**: Update TodoWrite after EVERY phase completion before starting next phase
-7. **Lightweight Index**: Phase 3 only generates SKILL.md index by reading existing files
+1. **Start Immediately**: First action is TodoWrite initialization
+2. **Path-Conditional Output**: Every rule file includes `paths` frontmatter
+3. **Template-Driven**: Agent reads templates before generating content
+4. **Agent Produces Files**: Agent writes all rule files directly
+5. **No Manual Loading**: Rules auto-activate when Claude works with matching files
 
 ---
 
 ## 3-Phase Execution
 
-### Phase 1: Prepare Context Paths
+### Phase 1: Prepare Context & Detect Tech Stack
 
-**Goal**: Detect input mode, prepare context paths for agent, check existing SKILL
+**Goal**: Detect input mode, extract tech stack info, determine file extensions
 
 **Input Mode Detection**:
 ```bash
-# Get input parameter
 input="$1"
 
-# Detect mode
 if [[ "$input" == WFS-* ]]; then
   MODE="session"
   SESSION_ID="$input"
-  CONTEXT_PATH=".workflow/${SESSION_ID}"
+  # Read workflow-session.json to extract tech stack
 else
   MODE="direct"
   TECH_STACK_NAME="$input"
-  CONTEXT_PATH="$input"  # Pass tech stack name as context
 fi
 ```
 
-**Check Existing SKILL**:
-```bash
-# For session mode, peek at session to get tech stack name
-if [[ "$MODE" == "session" ]]; then
-  bash(test -f ".workflow/${SESSION_ID}/workflow-session.json")
-  Read(.workflow/${SESSION_ID}/workflow-session.json)
-  # Extract tech_stack_name (minimal extraction)
-fi
+**Tech Stack Analysis**:
+```javascript
+// Decompose composite tech stacks
+// "typescript-react-nextjs" → ["typescript", "react", "nextjs"]
 
-# Normalize and check
+const TECH_EXTENSIONS = {
+  "typescript": "{ts,tsx}",
+  "javascript": "{js,jsx}",
+  "python": "py",
+  "rust": "rs",
+  "go": "go",
+  "java": "java",
+  "csharp": "cs",
+  "ruby": "rb",
+  "php": "php"
+};
+
+const FRAMEWORK_TYPE = {
+  "react": "frontend",
+  "vue": "frontend",
+  "angular": "frontend",
+  "nextjs": "fullstack",
+  "nuxt": "fullstack",
+  "fastapi": "backend",
+  "express": "backend",
+  "django": "backend",
+  "rails": "backend"
+};
+```
+
+**Check Existing Rules**:
+```bash
 normalized_name=$(echo "$TECH_STACK_NAME" | tr '[:upper:]' '[:lower:]' | tr ' ' '-')
-bash(test -d ".claude/skills/${normalized_name}" && echo "exists" || echo "not_exists")
-bash(find ".claude/skills/${normalized_name}" -name "*.md" 2>/dev/null | wc -l || echo 0)
+rules_dir=".claude/rules/tech/${normalized_name}"
+existing_count=$(find "${rules_dir}" -name "*.md" 2>/dev/null | wc -l || echo 0)
 ```
 
 **Skip Decision**:
-```javascript
-if (existing_files > 0 && !regenerate_flag) {
-  SKIP_GENERATION = true
-  message = "Tech stack SKILL already exists, skipping Phase 2. Use --regenerate to force regeneration."
-} else if (regenerate_flag) {
-  bash(rm -rf ".claude/skills/${normalized_name}")
-  SKIP_GENERATION = false
-  message = "Regenerating tech stack SKILL from scratch."
-} else {
-  SKIP_GENERATION = false
-  message = "No existing SKILL found, generating new tech stack documentation."
-}
-```
+- If `existing_count > 0` AND no `--regenerate` → `SKIP_GENERATION = true`
+- If `--regenerate` → Delete existing and regenerate
 
 **Output Variables**:
-- `MODE`: `session` or `direct`
-- `SESSION_ID`: Session ID (if session mode)
-- `CONTEXT_PATH`: Path to session directory OR tech stack name
-- `TECH_STACK_NAME`: Extracted or provided tech stack name
-- `SKIP_GENERATION`: Boolean - whether to skip Phase 2
+- `TECH_STACK_NAME`: Normalized name
+- `PRIMARY_LANG`: Primary language
+- `FILE_EXT`: File extension pattern
+- `FRAMEWORK_TYPE`: frontend | backend | fullstack | library
+- `COMPONENTS`: Array of tech components
+- `SKIP_GENERATION`: Boolean
 
-**TodoWrite**:
-- If skipping: Mark phase 1 completed, phase 2 completed, phase 3 in_progress
-- If not skipping: Mark phase 1 completed, phase 2 in_progress
+**TodoWrite**: Mark phase 1 completed
 
 ---
 
-### Phase 2: Agent Produces All Files
+### Phase 2: Agent Produces Path-Conditional Rules
 
 **Skip Condition**: Skipped if `SKIP_GENERATION = true`
 
-**Goal**: Delegate EVERYTHING to agent - context reading, Exa research, content synthesis, and file writing
-
-**Agent Task Specification**:
+**Goal**: Delegate to agent for Exa research and rule file generation
 
+**Template Files**:
 ```
-Task(
+~/.claude/workflows/cli-templates/prompts/rules/
+├── tech-rules-agent-prompt.txt  # Agent instructions
+├── rule-core.txt                # Core principles template
+├── rule-patterns.txt            # Implementation patterns template
+├── rule-testing.txt             # Testing rules template
+├── rule-config.txt              # Configuration rules template
+├── rule-api.txt                 # API rules template (backend)
+└── rule-components.txt          # Component rules template (frontend)
+```
+
+**Agent Task**:
+
+```javascript
+Task({
   subagent_type: "general-purpose",
-  description: "Generate tech stack SKILL: {CONTEXT_PATH}",
-  prompt: "
-Generate a complete tech stack SKILL package with Exa research.
+  description: `Generate tech stack rules: ${TECH_STACK_NAME}`,
+  prompt: `
+You are generating path-conditional rules for Claude Code.
 
-**Context Provided**:
-- Mode: {MODE}
-- Context Path: {CONTEXT_PATH}
+## Context
+- Tech Stack: ${TECH_STACK_NAME}
+- Primary Language: ${PRIMARY_LANG}
+- File Extensions: ${FILE_EXT}
+- Framework Type: ${FRAMEWORK_TYPE}
+- Components: ${JSON.stringify(COMPONENTS)}
+- Output Directory: .claude/rules/tech/${TECH_STACK_NAME}/
 
-**Templates Available**:
-- Module Format: ~/.claude/workflows/cli-templates/prompts/tech/tech-module-format.txt
-- SKILL Index: ~/.claude/workflows/cli-templates/prompts/tech/tech-skill-index.txt
+## Instructions
 
-**Your Responsibilities**:
+Read the agent prompt template for detailed instructions:
+$(cat ~/.claude/workflows/cli-templates/prompts/rules/tech-rules-agent-prompt.txt)
 
-1. **Extract Tech Stack Information**:
+## Execution Steps
 
-   IF MODE == 'session':
-     - Read `.workflow/active/{session_id}/workflow-session.json`
-     - Read `.workflow/active/{session_id}/.process/context-package.json`
-     - Extract tech_stack: {language, frameworks, libraries}
-     - Build tech stack name: \"{language}-{framework1}-{framework2}\"
-     - Example: \"typescript-react-nextjs\"
+1. Execute Exa research queries (see agent prompt)
+2. Read each rule template
+3. Generate rule files following template structure
+4. Write files to output directory
+5. Write metadata.json
+6. Report completion
 
-   IF MODE == 'direct':
-     - Tech stack name = CONTEXT_PATH
-     - Parse composite: split by '-' delimiter
-     - Example: \"typescript-react-nextjs\" → [\"typescript\", \"react\", \"nextjs\"]
+## Variable Substitutions
 
-2. **Execute Exa Research** (4-6 parallel queries):
-
-   Base Queries (always execute):
-   - mcp__exa__get_code_context_exa(query: \"{tech} core principles best practices 2025\", tokensNum: 8000)
-   - mcp__exa__get_code_context_exa(query: \"{tech} common patterns architecture examples\", tokensNum: 7000)
-   - mcp__exa__web_search_exa(query: \"{tech} configuration setup tooling 2025\", numResults: 5)
-   - mcp__exa__get_code_context_exa(query: \"{tech} testing strategies\", tokensNum: 5000)
-
-   Component Queries (if composite):
-   - For each additional component:
-     mcp__exa__get_code_context_exa(query: \"{main_tech} {component} integration\", tokensNum: 5000)
-
-3. **Read Module Format Template**:
-
-   Read template for structure guidance:
-   ```bash
-   Read(~/.claude/workflows/cli-templates/prompts/tech/tech-module-format.txt)
-   ```
-
-4. **Synthesize Content into 6 Modules**:
-
-   Follow template structure from tech-module-format.txt:
-   - **principles.md** - Core concepts, philosophies (~3K tokens)
-   - **patterns.md** - Implementation patterns with code examples (~5K tokens)
-   - **practices.md** - Best practices, anti-patterns, pitfalls (~4K tokens)
-   - **testing.md** - Testing strategies, frameworks (~3K tokens)
-   - **config.md** - Setup, configuration, tooling (~3K tokens)
-   - **frameworks.md** - Framework integration (only if composite, ~4K tokens)
-
-   Each module follows template format:
-   - Frontmatter (YAML)
-   - Main sections with clear headings
-   - Code examples from Exa research
-   - Best practices sections
-   - References to Exa sources
-
-5. **Write Files Directly**:
-
-   ```javascript
-   // Create directory
-   bash(mkdir -p \".claude/skills/{tech_stack_name}\")
-
-   // Write each module file using Write tool
-   Write({ file_path: \".claude/skills/{tech_stack_name}/principles.md\", content: ... })
-   Write({ file_path: \".claude/skills/{tech_stack_name}/patterns.md\", content: ... })
-   Write({ file_path: \".claude/skills/{tech_stack_name}/practices.md\", content: ... })
-   Write({ file_path: \".claude/skills/{tech_stack_name}/testing.md\", content: ... })
-   Write({ file_path: \".claude/skills/{tech_stack_name}/config.md\", content: ... })
-   // Write frameworks.md only if composite
-
-   // Write metadata.json
-   Write({
-     file_path: \".claude/skills/{tech_stack_name}/metadata.json\",
-     content: JSON.stringify({
-       tech_stack_name,
-       components,
-       is_composite,
-       generated_at: timestamp,
-       source: \"exa-research\",
-       research_summary: { total_queries, total_sources }
-     })
-   })
-   ```
-
-6. **Report Completion**:
-
-   Provide summary:
-   - Tech stack name
-   - Files created (count)
-   - Exa queries executed
-   - Sources consulted
-
-**CRITICAL**:
-- MUST read external template files before generating content (step 3 for modules, step 4 for index)
-- You have FULL autonomy - read files, execute Exa, synthesize content, write files
-- Do NOT return JSON or structured data - produce actual .md files
-- Handle errors gracefully (Exa failures, missing files, template read failures)
-- If tech stack cannot be determined, ask orchestrator to clarify
-  "
-)
+Replace in templates:
+- {TECH_STACK_NAME} → ${TECH_STACK_NAME}
+- {PRIMARY_LANG} → ${PRIMARY_LANG}
+- {FILE_EXT} → ${FILE_EXT}
+- {FRAMEWORK_TYPE} → ${FRAMEWORK_TYPE}
+`
+})
 ```
 
 **Completion Criteria**:
-- Agent task executed successfully
-- 5-6 modular files written to `.claude/skills/{tech_stack_name}/`
+- 4-6 rule files written with proper `paths` frontmatter
 - metadata.json written
-- Agent reports completion
+- Agent reports files created
 
-**TodoWrite**: Mark phase 2 completed, phase 3 in_progress
+**TodoWrite**: Mark phase 2 completed
 
 ---
 
-### Phase 3: Generate SKILL.md Index
+### Phase 3: Verify & Report
 
-**Note**: This phase **ALWAYS executes** - generates or updates the SKILL index.
-
-**Goal**: Read generated module files and create SKILL.md index with loading recommendations
+**Goal**: Verify generated files and provide usage summary
 
 **Steps**:
 
-1. **Verify Generated Files**:
+1. **Verify Files**:
    ```bash
-   bash(find ".claude/skills/${TECH_STACK_NAME}" -name "*.md" -type f | sort)
+   find ".claude/rules/tech/${TECH_STACK_NAME}" -name "*.md" -type f
    ```
 
-2. **Read metadata.json**:
+2. **Validate Frontmatter**:
+   ```bash
+   head -5 ".claude/rules/tech/${TECH_STACK_NAME}/core.md"
+   ```
+
+3. **Read Metadata**:
    ```javascript
-   Read(.claude/skills/${TECH_STACK_NAME}/metadata.json)
-   // Extract: tech_stack_name, components, is_composite, research_summary
+   Read(`.claude/rules/tech/${TECH_STACK_NAME}/metadata.json`)
    ```
 
-3. **Read Module Headers** (optional, first 20 lines):
-   ```javascript
-   Read(.claude/skills/${TECH_STACK_NAME}/principles.md, limit: 20)
-   // Repeat for other modules
+4. **Generate Summary Report**:
    ```
+   Tech Stack Rules Generated
 
-4. **Read SKILL Index Template**:
+   Tech Stack: {TECH_STACK_NAME}
+   Location: .claude/rules/tech/{TECH_STACK_NAME}/
 
-   ```javascript
-   Read(~/.claude/workflows/cli-templates/prompts/tech/tech-skill-index.txt)
+   Files Created:
+   ├── core.md         → paths: **/*.{ext}
+   ├── patterns.md     → paths: src/**/*.{ext}
+   ├── testing.md      → paths: **/*.{test,spec}.{ext}
+   ├── config.md       → paths: *.config.*
+   ├── api.md          → paths: **/api/**/* (if backend)
+   └── components.md   → paths: **/components/**/* (if frontend)
+
+   Auto-Loading:
+   - Rules apply automatically when editing matching files
+   - No manual loading required
+
+   Example Activation:
+   - Edit src/components/Button.tsx → core.md + patterns.md + components.md
+   - Edit tests/api.test.ts → core.md + testing.md
+   - Edit package.json → config.md
    ```
 
-5. **Generate SKILL.md Index**:
-
-   Follow template from tech-skill-index.txt with variable substitutions:
-   - `{TECH_STACK_NAME}`: From metadata.json
-   - `{MAIN_TECH}`: Primary technology
-   - `{ISO_TIMESTAMP}`: Current timestamp
-   - `{QUERY_COUNT}`: From research_summary
-   - `{SOURCE_COUNT}`: From research_summary
-   - Conditional sections for composite tech stacks
-
-   Template provides structure for:
-   - Frontmatter with metadata
-   - Overview and tech stack description
-   - Module organization (Core/Practical/Config sections)
-   - Loading recommendations (Quick/Implementation/Complete)
-   - Usage guidelines and auto-trigger keywords
-   - Research metadata and version history
-
-6. **Write SKILL.md**:
-   ```javascript
-   Write({
-     file_path: `.claude/skills/${TECH_STACK_NAME}/SKILL.md`,
-     content: generatedIndexMarkdown
-   })
-   ```
-
-**Completion Criteria**:
-- SKILL.md index written
-- All module files verified
-- Loading recommendations included
-
 **TodoWrite**: Mark phase 3 completed
 
-**Final Report**:
-```
-Tech Stack SKILL Package Complete
-
-Tech Stack: {TECH_STACK_NAME}
-Location: .claude/skills/{TECH_STACK_NAME}/
-
-Files: SKILL.md + 5-6 modules + metadata.json
-Exa Research: {queries} queries, {sources} sources
-
-Usage: Skill(command: "{TECH_STACK_NAME}")
-```
-
 ---
 
-## Implementation Details
+## Path Pattern Reference
 
-### TodoWrite Patterns
+| Pattern | Matches |
+|---------|---------|
+| `**/*.ts` | All .ts files |
+| `src/**/*` | All files under src/ |
+| `*.config.*` | Config files in root |
+| `**/*.{ts,tsx}` | .ts and .tsx files |
 
-**Initialization** (Before Phase 1):
-```javascript
-TodoWrite({todos: [
-  {"content": "Prepare context paths", "status": "in_progress", "activeForm": "Preparing context paths"},
-  {"content": "Agent produces all module files", "status": "pending", "activeForm": "Agent producing files"},
-  {"content": "Generate SKILL.md index", "status": "pending", "activeForm": "Generating SKILL index"}
-]})
-```
-
-**Full Path** (SKIP_GENERATION = false):
-```javascript
-// After Phase 1
-TodoWrite({todos: [
-  {"content": "Prepare context paths", "status": "completed", ...},
-  {"content": "Agent produces all module files", "status": "in_progress", ...},
-  {"content": "Generate SKILL.md index", "status": "pending", ...}
-]})
-
-// After Phase 2
-TodoWrite({todos: [
-  {"content": "Prepare context paths", "status": "completed", ...},
-  {"content": "Agent produces all module files", "status": "completed", ...},
-  {"content": "Generate SKILL.md index", "status": "in_progress", ...}
-]})
-
-// After Phase 3
-TodoWrite({todos: [
-  {"content": "Prepare context paths", "status": "completed", ...},
-  {"content": "Agent produces all module files", "status": "completed", ...},
-  {"content": "Generate SKILL.md index", "status": "completed", ...}
-]})
-```
-
-**Skip Path** (SKIP_GENERATION = true):
-```javascript
-// After Phase 1 (skip Phase 2)
-TodoWrite({todos: [
-  {"content": "Prepare context paths", "status": "completed", ...},
-  {"content": "Agent produces all module files", "status": "completed", ...},  // Skipped
-  {"content": "Generate SKILL.md index", "status": "in_progress", ...}
-]})
-```
-
-### Execution Flow
-
-**Full Path**:
-```
-User → TodoWrite Init → Phase 1 (prepare) → Phase 2 (agent writes files) → Phase 3 (write index) → Report
-```
-
-**Skip Path**:
-```
-User → TodoWrite Init → Phase 1 (detect existing) → Phase 3 (update index) → Report
-```
-
-### Error Handling
-
-**Phase 1 Errors**:
-- Invalid session ID: Report error, verify session exists
-- Missing context-package: Warn, fall back to direct mode
-- No tech stack detected: Ask user to specify tech stack name
-
-**Phase 2 Errors (Agent)**:
-- Agent task fails: Retry once, report if fails again
-- Exa API failures: Agent handles internally with retries
-- Incomplete results: Warn user, proceed with partial data if minimum sections available
-
-**Phase 3 Errors**:
-- Write failures: Report which files failed
-- Missing files: Note in SKILL.md, suggest regeneration
+| Tech Stack | Core Pattern | Test Pattern |
+|------------|--------------|--------------|
+| TypeScript | `**/*.{ts,tsx}` | `**/*.{test,spec}.{ts,tsx}` |
+| Python | `**/*.py` | `**/test_*.py, **/*_test.py` |
+| Rust | `**/*.rs` | `**/tests/**/*.rs` |
+| Go | `**/*.go` | `**/*_test.go` |
 
 ---
 
 ## Parameters
 
 ```bash
-/memory:tech-research [session-id | "tech-stack-name"] [--regenerate] [--tool <gemini|qwen>]
+/memory:tech-research [session-id | "tech-stack-name"] [--regenerate]
 ```
 
 **Arguments**:
-- **session-id | tech-stack-name**: Input source (auto-detected by WFS- prefix)
-  - Session mode: `WFS-user-auth-v2` - Extract tech stack from workflow
-  - Direct mode: `"typescript"`, `"typescript-react-nextjs"` - User specifies
-- **--regenerate**: Force regenerate existing SKILL (deletes and recreates)
-- **--tool**: Reserved for future CLI integration (default: gemini)
+- **session-id**: `WFS-*` format - Extract from workflow session
+- **tech-stack-name**: Direct input - `"typescript"`, `"typescript-react"`
+- **--regenerate**: Force regenerate existing rules
 
 ---
 
 ## Examples
 
-**Generated File Structure** (for all examples):
-```
-.claude/skills/{tech-stack}/
-├── SKILL.md           # Index (Phase 3)
-├── principles.md      # Agent (Phase 2)
-├── patterns.md        # Agent
-├── practices.md       # Agent
-├── testing.md         # Agent
-├── config.md          # Agent
-├── frameworks.md      # Agent (if composite)
-└── metadata.json      # Agent
-```
-
-### Direct Mode - Single Stack
+### Single Language
 
 ```bash
 /memory:tech-research "typescript"
 ```
 
-**Workflow**:
-1. Phase 1: Detects direct mode, checks existing SKILL
-2. Phase 2: Agent executes 4 Exa queries, writes 5 modules
-3. Phase 3: Generates SKILL.md index
+**Output**: `.claude/rules/tech/typescript/` with 4 rule files
 
-### Direct Mode - Composite Stack
+### Frontend Stack
 
 ```bash
-/memory:tech-research "typescript-react-nextjs"
+/memory:tech-research "typescript-react"
 ```
 
-**Workflow**:
-1. Phase 1: Decomposes into ["typescript", "react", "nextjs"]
-2. Phase 2: Agent executes 6 Exa queries (4 base + 2 components), writes 6 modules (adds frameworks.md)
-3. Phase 3: Generates SKILL.md index with framework integration
+**Output**: `.claude/rules/tech/typescript-react/` with 5 rule files (includes components.md)
 
-### Session Mode - Extract from Workflow
+### Backend Stack
+
+```bash
+/memory:tech-research "python-fastapi"
+```
+
+**Output**: `.claude/rules/tech/python-fastapi/` with 5 rule files (includes api.md)
+
+### From Session
 
 ```bash
 /memory:tech-research WFS-user-auth-20251104
 ```
 
-**Workflow**:
-1. Phase 1: Reads session, extracts tech stack: `python-fastapi-sqlalchemy`
-2. Phase 2: Agent researches Python + FastAPI + SQLAlchemy, writes 6 modules
-3. Phase 3: Generates SKILL.md index
+**Workflow**: Extract tech stack from session → Generate rules
 
-### Regenerate Existing
+---
 
-```bash
-/memory:tech-research "react" --regenerate
-```
-
-**Workflow**:
-1. Phase 1: Deletes existing SKILL due to --regenerate
-2. Phase 2: Agent executes fresh Exa research (latest 2025 practices)
-3. Phase 3: Generates updated SKILL.md
-
-### Skip Path - Fast Update
-
-```bash
-/memory:tech-research "python"
-```
-
-**Scenario**: SKILL already exists with 7 files
-
-**Workflow**:
-1. Phase 1: Detects existing SKILL, sets SKIP_GENERATION = true
-2. Phase 2: **SKIPPED**
-3. Phase 3: Updates SKILL.md index only (5-10x faster)
+## Comparison: Rules vs SKILL
 
+| Aspect | SKILL Memory | Rules |
+|--------|--------------|-------|
+| Loading | Manual: `Skill("tech")` | Automatic by path |
+| Scope | All files when loaded | Only matching files |
+| Granularity | Monolithic packages | Per-file-type |
+| Context | Full package | Only relevant rules |
 
+**When to Use**:
+- **Rules**: Tech stack conventions per file type
+- **SKILL**: Reference docs, APIs, examples for manual lookup
diff --git a/.claude/skills/command-guide/reference/commands/memory/workflow-skill-memory.md b/.claude/skills/command-guide/reference/commands/memory/workflow-skill-memory.md
index c23b7c97..96be94b5 100644
--- a/.claude/skills/command-guide/reference/commands/memory/workflow-skill-memory.md
+++ b/.claude/skills/command-guide/reference/commands/memory/workflow-skill-memory.md
@@ -187,7 +187,7 @@ Objectives:
 
 3. Use Gemini for aggregation (optional):
    Command pattern:
-   cd .workflow/.archives/{session_id} && gemini -p "
+   ccw cli exec "
    PURPOSE: Extract lessons and conflicts from workflow session
    TASK:
    • Analyze IMPL_PLAN and lessons from manifest
@@ -198,7 +198,7 @@ Objectives:
    CONTEXT: @IMPL_PLAN.md @workflow-session.json
    EXPECTED: Structured lessons and conflicts in JSON format
    RULES: Template reference from skill-aggregation.txt
-   "
+   " --tool gemini --cd .workflow/.archives/{session_id}
 
 3.5. **Generate SKILL.md Description** (CRITICAL for auto-loading):
 
@@ -334,7 +334,7 @@ Objectives:
    - Sort sessions by date
 
 2. Use Gemini for final aggregation:
-   gemini -p "
+   ccw cli exec "
    PURPOSE: Aggregate lessons and conflicts from all workflow sessions
    TASK:
    • Group successes by functional domain
@@ -345,7 +345,7 @@ Objectives:
    CONTEXT: [Provide aggregated JSON data]
    EXPECTED: Final aggregated structure for SKILL documents
    RULES: Template reference from skill-aggregation.txt
-   "
+   " --tool gemini
 
 3. Read templates for formatting (same 4 templates as single mode)
 
diff --git a/.claude/skills/command-guide/reference/commands/workflow/execute.md b/.claude/skills/command-guide/reference/commands/workflow/execute.md
index 669622a9..97465596 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/execute.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/execute.md
@@ -67,7 +67,9 @@ Phase 4: Execution Strategy & Task Execution
          ├─ Get next in_progress task from TodoWrite
          ├─ Lazy load task JSON
          ├─ Launch agent with task context
-         ├─ Mark task completed
+         ├─ Mark task completed (update IMPL-*.json status)
+         │  # Update task status with ccw session (auto-tracks status_history):
+         │  # ccw session task ${sessionId} IMPL-X completed
          └─ Advance to next task
 
 Phase 5: Completion
@@ -90,37 +92,32 @@ Resume Mode (--resume-session):
 
 **Process**:
 
-#### Step 1.1: Count Active Sessions
+#### Step 1.1: List Active Sessions
 ```bash
-bash(find .workflow/active/ -name "WFS-*" -type d 2>/dev/null | wc -l)
+ccw session list --location active
+# Returns: {"success":true,"result":{"active":[...],"total":N}}
 ```
 
 #### Step 1.2: Handle Session Selection
 
-**Case A: No Sessions** (count = 0)
+**Case A: No Sessions** (total = 0)
 ```
 ERROR: No active workflow sessions found
 Run /workflow:plan "task description" to create a session
 ```
 
-**Case B: Single Session** (count = 1)
-```bash
-bash(find .workflow/active/ -name "WFS-*" -type d 2>/dev/null | head -1 | xargs basename)
-```
-Auto-select and continue to Phase 2.
+**Case B: Single Session** (total = 1)
+Auto-select the single session from result.active[0].session_id and continue to Phase 2.
 
-**Case C: Multiple Sessions** (count > 1)
+**Case C: Multiple Sessions** (total > 1)
 
-List sessions with metadata and prompt user selection:
+List sessions with metadata using ccw session:
 ```bash
-bash(for dir in .workflow/active/WFS-*/; do
-  session=$(basename "$dir")
-  project=$(ccw session read "$session" --type session --raw 2>/dev/null | jq -r '.project // "Unknown"')
-  total=$(grep -c "^- \[" "$dir/TODO_LIST.md" 2>/dev/null || echo "0")
-  completed=$(grep -c "^- \[x\]" "$dir/TODO_LIST.md" 2>/dev/null || echo "0")
-  [ "$total" -gt 0 ] && progress=$((completed * 100 / total)) || progress=0
-  echo "${session} | ${project} | ${completed}/${total} tasks (${progress}%)"
-done)
+# Get session list with metadata
+ccw session list --location active
+
+# For each session, get stats
+ccw session stats WFS-session-name
 ```
 
 Use AskUserQuestion to present formatted options (max 4 options shown):
@@ -158,6 +155,14 @@ ccw session read ${sessionId} --type session
 **Output**: Store session metadata in memory
 **DO NOT read task JSONs yet** - defer until execution phase (lazy loading)
 
+#### Step 1.4: Update Session Status to Active
+**Purpose**: Update workflow-session.json status from "planning" to "active" for dashboard monitoring.
+
+```bash
+# Update status atomically using ccw session
+ccw session status ${sessionId} active
+```
+
 **Resume Mode**: This entire phase is skipped when `--resume-session="session-id"` flag is provided.
 
 ### Phase 2: Planning Document Validation
diff --git a/.claude/skills/command-guide/reference/commands/workflow/lite-execute.md b/.claude/skills/command-guide/reference/commands/workflow/lite-execute.md
index 21f46ec2..a408b855 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/lite-execute.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/lite-execute.md
@@ -473,7 +473,7 @@ Detailed plan: ${executionContext.session.artifacts.plan}`)
   return prompt
 }
 
-codex --full-auto exec "${buildCLIPrompt(batch)}" --skip-git-repo-check -s danger-full-access
+ccw cli exec "${buildCLIPrompt(batch)}" --tool codex --mode auto
 ```
 
 **Execution with tracking**:
@@ -541,15 +541,15 @@ RULES: $(cat ~/.claude/workflows/cli-templates/prompts/analysis/02-review-code-q
 # - Report findings directly
 
 # Method 2: Gemini Review (recommended)
-gemini -p "[Shared Prompt Template with artifacts]"
+ccw cli exec "[Shared Prompt Template with artifacts]" --tool gemini
 # CONTEXT includes: @**/* @${plan.json} [@${exploration.json}]
 
 # Method 3: Qwen Review (alternative)
-qwen -p "[Shared Prompt Template with artifacts]"
+ccw cli exec "[Shared Prompt Template with artifacts]" --tool qwen
 # Same prompt as Gemini, different execution engine
 
 # Method 4: Codex Review (autonomous)
-codex --full-auto exec "[Verify plan acceptance criteria at ${plan.json}]" --skip-git-repo-check -s danger-full-access
+ccw cli exec "[Verify plan acceptance criteria at ${plan.json}]" --tool codex --mode auto
 ```
 
 **Implementation Note**: Replace `[Shared Prompt Template with artifacts]` placeholder with actual template content, substituting:
diff --git a/.claude/skills/command-guide/reference/commands/workflow/lite-plan.md b/.claude/skills/command-guide/reference/commands/workflow/lite-plan.md
index 1c8822c8..d575869d 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/lite-plan.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/lite-plan.md
@@ -152,11 +152,16 @@ Launching ${selectedAngles.length} parallel explorations...
 
 **Launch Parallel Explorations** - Orchestrator assigns angle to each agent:
 
+**⚠️ CRITICAL - NO BACKGROUND EXECUTION**:
+- **MUST NOT use `run_in_background: true`** - exploration results are REQUIRED before planning
+
+
 ```javascript
 // Launch agents with pre-assigned angles
 const explorationTasks = selectedAngles.map((angle, index) =>
   Task(
     subagent_type="cli-explore-agent",
+    run_in_background=false,  // ⚠️ MANDATORY: Must wait for results
     description=`Explore: ${angle}`,
     prompt=`
 ## Task Objective
@@ -356,7 +361,15 @@ if (dedupedClarifications.length > 0) {
 // Step 1: Read schema
 const schema = Bash(`cat ~/.claude/workflows/cli-templates/schemas/plan-json-schema.json`)
 
-// Step 2: Generate plan following schema (Claude directly, no agent)
+// Step 2: ⚠️ MANDATORY - Read and review ALL exploration files
+const manifest = JSON.parse(Read(`${sessionFolder}/explorations-manifest.json`))
+manifest.explorations.forEach(exp => {
+  const explorationData = Read(exp.path)
+  console.log(`\n### Exploration: ${exp.angle}\n${explorationData}`)
+})
+
+// Step 3: Generate plan following schema (Claude directly, no agent)
+// ⚠️ Plan MUST incorporate insights from exploration files read in Step 2
 const plan = {
   summary: "...",
   approach: "...",
@@ -367,10 +380,10 @@ const plan = {
   _metadata: { timestamp: getUtc8ISOString(), source: "direct-planning", planning_mode: "direct" }
 }
 
-// Step 3: Write plan to session folder
+// Step 4: Write plan to session folder
 Write(`${sessionFolder}/plan.json`, JSON.stringify(plan, null, 2))
 
-// Step 4: MUST continue to Phase 4 (Confirmation) - DO NOT execute code here
+// Step 5: MUST continue to Phase 4 (Confirmation) - DO NOT execute code here
 ```
 
 **Medium/High Complexity** - Invoke cli-lite-planning-agent:
diff --git a/.claude/skills/command-guide/reference/commands/workflow/plan.md b/.claude/skills/command-guide/reference/commands/workflow/plan.md
index 938fc719..f2b4b611 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/plan.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/plan.md
@@ -61,7 +61,7 @@ Phase 2: Context Gathering
       ├─ Tasks attached: Analyze structure → Identify integration → Generate package
       └─ Output: contextPath + conflict_risk
 
-Phase 3: Conflict Resolution (conditional)
+Phase 3: Conflict Resolution 
    └─ Decision (conflict_risk check):
       ├─ conflict_risk ≥ medium → Execute /workflow:tools:conflict-resolution
       │   ├─ Tasks attached: Detect conflicts → Present to user → Apply strategies
@@ -168,7 +168,7 @@ SlashCommand(command="/workflow:tools:context-gather --session [sessionId] \"[st
 
 ---
 
-### Phase 3: Conflict Resolution (Optional - auto-triggered by conflict risk)
+### Phase 3: Conflict Resolution
 
 **Trigger**: Only execute when context-package.json indicates conflict_risk is "medium" or "high"
 
@@ -185,10 +185,10 @@ SlashCommand(command="/workflow:tools:conflict-resolution --session [sessionId]
 
 **Parse Output**:
 - Extract: Execution status (success/skipped/failed)
-- Verify: CONFLICT_RESOLUTION.md file path (if executed)
+- Verify: conflict-resolution.json file path (if executed)
 
 **Validation**:
-- File `.workflow/active/[sessionId]/.process/CONFLICT_RESOLUTION.md` exists (if executed)
+- File `.workflow/active/[sessionId]/.process/conflict-resolution.json` exists (if executed)
 
 **Skip Behavior**:
 - If conflict_risk is "none" or "low", skip directly to Phase 3.5
@@ -497,7 +497,7 @@ Return summary to user
 - Parse context path from Phase 2 output, store in memory
 - **Extract conflict_risk from context-package.json**: Determine Phase 3 execution
 - **If conflict_risk ≥ medium**: Launch Phase 3 conflict-resolution with sessionId and contextPath
-- Wait for Phase 3 to finish executing (if executed), verify CONFLICT_RESOLUTION.md created
+- Wait for Phase 3 to finish executing (if executed), verify conflict-resolution.json created
 - **If conflict_risk is none/low**: Skip Phase 3, proceed directly to Phase 4
 - **Build Phase 4 command**: `/workflow:tools:task-generate-agent --session [sessionId]`
 - Pass session ID to Phase 4 command
diff --git a/.claude/skills/command-guide/reference/commands/workflow/review.md b/.claude/skills/command-guide/reference/commands/workflow/review.md
index e40795a9..0d2c2e4f 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/review.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/review.md
@@ -133,37 +133,37 @@ After bash validation, the model takes control to:
      ```
    - Use Gemini for security analysis:
      ```bash
-     cd .workflow/active/${sessionId} && gemini -p "
+     ccw cli exec "
      PURPOSE: Security audit of completed implementation
      TASK: Review code for security vulnerabilities, insecure patterns, auth/authz issues
      CONTEXT: @.summaries/IMPL-*.md,../.. @../../CLAUDE.md
      EXPECTED: Security findings report with severity levels
      RULES: Focus on OWASP Top 10, authentication, authorization, data validation, injection risks
-     " --approval-mode yolo
+     " --tool gemini --mode write --cd .workflow/active/${sessionId}
      ```
 
    **Architecture Review** (`--type=architecture`):
    - Use Qwen for architecture analysis:
      ```bash
-     cd .workflow/active/${sessionId} && qwen -p "
+     ccw cli exec "
      PURPOSE: Architecture compliance review
      TASK: Evaluate adherence to architectural patterns, identify technical debt, review design decisions
      CONTEXT: @.summaries/IMPL-*.md,../.. @../../CLAUDE.md
      EXPECTED: Architecture assessment with recommendations
      RULES: Check for patterns, separation of concerns, modularity, scalability
-     " --approval-mode yolo
+     " --tool qwen --mode write --cd .workflow/active/${sessionId}
      ```
 
    **Quality Review** (`--type=quality`):
    - Use Gemini for code quality:
      ```bash
-     cd .workflow/active/${sessionId} && gemini -p "
+     ccw cli exec "
      PURPOSE: Code quality and best practices review
      TASK: Assess code readability, maintainability, adherence to best practices
      CONTEXT: @.summaries/IMPL-*.md,../.. @../../CLAUDE.md
      EXPECTED: Quality assessment with improvement suggestions
      RULES: Check for code smells, duplication, complexity, naming conventions
-     " --approval-mode yolo
+     " --tool gemini --mode write --cd .workflow/active/${sessionId}
      ```
 
    **Action Items Review** (`--type=action-items`):
@@ -177,7 +177,7 @@ After bash validation, the model takes control to:
      '
 
      # Check implementation summaries against requirements
-     cd .workflow/active/${sessionId} && gemini -p "
+     ccw cli exec "
      PURPOSE: Verify all requirements and acceptance criteria are met
      TASK: Cross-check implementation summaries against original requirements
      CONTEXT: @.task/IMPL-*.json,.summaries/IMPL-*.md,../.. @../../CLAUDE.md
@@ -191,7 +191,7 @@ After bash validation, the model takes control to:
      - Verify all acceptance criteria are met
      - Flag any incomplete or missing action items
      - Assess deployment readiness
-     " --approval-mode yolo
+     " --tool gemini --mode write --cd .workflow/active/${sessionId}
      ```
 
 
diff --git a/.claude/skills/command-guide/reference/commands/workflow/session/complete.md b/.claude/skills/command-guide/reference/commands/workflow/session/complete.md
index 369a8107..96501dab 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/session/complete.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/session/complete.md
@@ -25,11 +25,9 @@ Mark the currently active workflow session as complete, analyze it for lessons l
 
 #### Step 1.1: Find Active Session and Get Name
 ```bash
-# Find active session directory
-bash(find .workflow/active/ -name "WFS-*" -type d | head -1)
-
-# Extract session name from directory path
-bash(basename .workflow/active/WFS-session-name)
+# Find active session
+ccw session list --location active
+# Extract first session_id from result.active array
 ```
 **Output**: Session name `WFS-session-name`
 
@@ -199,7 +197,7 @@ manifest.push(archiveEntry);
 Write('.workflow/archives/manifest.json', JSON.stringify(manifest, null, 2));
 ```
 
-#### Step 3.3: Remove Archiving Marker
+#### Step 3.5: Remove Archiving Marker
 ```bash
 # Remove archiving marker from archived session (use bash rm as ccw has no delete)
 rm .workflow/archives/WFS-session-name/.process/.archiving 2>/dev/null || true
@@ -223,7 +221,8 @@ rm .workflow/archives/WFS-session-name/.process/.archiving 2>/dev/null || true
 
 #### Step 4.1: Check Project State Exists
 ```bash
-bash(test -f .workflow/project.json && echo "EXISTS" || echo "SKIP")
+# Check project state using ccw
+ccw session read project --type project 2>/dev/null && echo "EXISTS" || echo "SKIP"
 ```
 
 **If SKIP**: Output warning and skip Phase 4
@@ -251,8 +250,8 @@ const featureId = title.toLowerCase().replace(/[^a-z0-9]+/g, '-').substring(0, 5
 #### Step 4.3: Update project.json
 
 ```bash
-# Read current project state
-bash(cat .workflow/project.json)
+# Read current project state using ccw
+ccw session read project --type project --raw
 ```
 
 **JSON Update Logic**:
@@ -366,8 +365,8 @@ function getLatestCommitHash() {
 **Recovery Steps**:
 ```bash
 # Session still in .workflow/active/WFS-session-name
-# Remove archiving marker
-bash(rm .workflow/active/WFS-session-name/.archiving)
+# Remove archiving marker using bash
+rm .workflow/active/WFS-session-name/.process/.archiving 2>/dev/null || true
 ```
 
 **User Notification**:
@@ -464,11 +463,12 @@ Session state: PARTIALLY COMPLETE (session archived, manifest needs update)
 
 **Phase 3: Atomic Commit** (Transactional file operations)
 - Create archive directory
+- Update session status to "completed"
 - Move session to archive location
 - Update manifest.json with archive entry
 - Remove `.archiving` marker
 - **All-or-nothing**: Either all succeed or session remains in safe state
-- **Total**: 4 bash commands + JSON manipulation
+- **Total**: 5 bash commands + JSON manipulation
 
 **Phase 4: Project Registry Update** (Optional feature tracking)
 - Check project.json exists
@@ -498,3 +498,55 @@ Session state: PARTIALLY COMPLETE (session archived, manifest needs update)
 - Idempotent operations (safe to retry)
 
 
+
+## session_manager Tool Alternative
+
+Use `ccw tool exec session_manager` for session completion operations:
+
+### List Active Sessions
+```bash
+ccw tool exec session_manager '{"operation":"list","location":"active"}'
+```
+
+### Update Session Status to Completed
+```bash
+ccw tool exec session_manager '{
+  "operation": "update",
+  "session_id": "WFS-xxx",
+  "content_type": "session",
+  "content": {
+    "status": "completed",
+    "archived_at": "2025-12-10T08:00:00Z"
+  }
+}'
+```
+
+### Archive Session
+```bash
+ccw tool exec session_manager '{"operation":"archive","session_id":"WFS-xxx"}'
+
+# This operation:
+# 1. Updates status to "completed" if update_status=true (default)
+# 2. Moves session from .workflow/active/ to .workflow/archives/
+```
+
+### Read Session Data
+```bash
+# Read workflow-session.json
+ccw tool exec session_manager '{"operation":"read","session_id":"WFS-xxx","content_type":"session"}'
+
+# Read IMPL_PLAN.md
+ccw tool exec session_manager '{"operation":"read","session_id":"WFS-xxx","content_type":"plan"}'
+```
+
+### Write Archiving Marker
+```bash
+ccw tool exec session_manager '{
+  "operation": "write",
+  "session_id": "WFS-xxx",
+  "content_type": "process",
+  "path_params": {"filename": ".archiving"},
+  "content": ""
+}'
+```
+
diff --git a/.claude/skills/command-guide/reference/commands/workflow/session/list.md b/.claude/skills/command-guide/reference/commands/workflow/session/list.md
index 34cbeea7..6ba221a3 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/session/list.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/session/list.md
@@ -17,41 +17,30 @@ Display all workflow sessions with their current status, progress, and metadata.
 
 ## Implementation Flow
 
-### Step 1: Find All Sessions
+### Step 1: List All Sessions
 ```bash
-ls .workflow/active/WFS-* 2>/dev/null
+ccw session list --location both
 ```
 
-### Step 2: Check Active Session
+### Step 2: Get Session Statistics
 ```bash
-find .workflow/active/ -name "WFS-*" -type d 2>/dev/null | head -1
+ccw session stats WFS-session
+# Returns: tasks count by status, summaries count, has_plan
 ```
 
 ### Step 3: Read Session Metadata
 ```bash
-jq -r '.session_id, .status, .project' .workflow/active/WFS-session/workflow-session.json
+ccw session read WFS-session --type session
+# Returns: session_id, status, project, created_at, etc.
 ```
 
-### Step 4: Count Task Progress
-```bash
-find .workflow/active/WFS-session/.task/ -name "*.json" -type f 2>/dev/null | wc -l
-find .workflow/active/WFS-session/.summaries/ -name "*.md" -type f 2>/dev/null | wc -l
-```
-
-### Step 5: Get Creation Time
-```bash
-jq -r '.created_at // "unknown"' .workflow/active/WFS-session/workflow-session.json
-```
-
-## Simple Bash Commands
+## Simple Commands
 
 ### Basic Operations
-- **List sessions**: `find .workflow/active/ -name "WFS-*" -type d`
-- **Find active**: `find .workflow/active/ -name "WFS-*" -type d`
-- **Read session data**: `jq -r '.session_id, .status' session.json`
-- **Count tasks**: `find .task/ -name "*.json" -type f | wc -l`
-- **Count completed**: `find .summaries/ -name "*.md" -type f 2>/dev/null | wc -l`
-- **Get timestamp**: `jq -r '.created_at' session.json`
+- **List all sessions**: `ccw session list`
+- **List active only**: `ccw session list --location active`
+- **Read session data**: `ccw session read WFS-xxx --type session`
+- **Get task stats**: `ccw session stats WFS-xxx`
 
 ## Simple Output Format
 
@@ -94,4 +83,32 @@ ccw session list --location active --no-metadata
 
 # Show recent sessions
 ccw session list --location active
-```
\ No newline at end of file
+```
+## session_manager Tool Alternative
+
+Use `ccw tool exec session_manager` for simplified session listing:
+
+### List All Sessions (Active + Archived)
+```bash
+ccw tool exec session_manager '{"operation":"list","location":"both","include_metadata":true}'
+
+# Response:
+# {
+#   "success": true,
+#   "result": {
+#     "active": [{"session_id":"WFS-xxx","metadata":{...}}],
+#     "archived": [{"session_id":"WFS-yyy","metadata":{...}}],
+#     "total": 2
+#   }
+# }
+```
+
+### List Active Sessions Only
+```bash
+ccw tool exec session_manager '{"operation":"list","location":"active","include_metadata":true}'
+```
+
+### Read Specific Session
+```bash
+ccw tool exec session_manager '{"operation":"read","session_id":"WFS-xxx","content_type":"session"}'
+```
diff --git a/.claude/skills/command-guide/reference/commands/workflow/session/resume.md b/.claude/skills/command-guide/reference/commands/workflow/session/resume.md
index 65ca4c0e..6780cef2 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/session/resume.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/session/resume.md
@@ -17,39 +17,35 @@ Resume the most recently paused workflow session, restoring all context and stat
 
 ### Step 1: Find Paused Sessions
 ```bash
-ls .workflow/active/WFS-* 2>/dev/null
+ccw session list --location active
+# Filter for sessions with status="paused"
 ```
 
 ### Step 2: Check Session Status
 ```bash
-jq -r '.status' .workflow/active/WFS-session/workflow-session.json
+ccw session read WFS-session --type session
+# Check .status field in response
 ```
 
 ### Step 3: Find Most Recent Paused
 ```bash
-ls -t .workflow/active/WFS-*/workflow-session.json | head -1
+ccw session list --location active
+# Sort by created_at, filter for paused status
 ```
 
-### Step 4: Update Session Status
+### Step 4: Update Session Status to Active
 ```bash
-jq '.status = "active"' .workflow/active/WFS-session/workflow-session.json > temp.json
-mv temp.json .workflow/active/WFS-session/workflow-session.json
+ccw session status WFS-session active
+# Or with full update:
+ccw session update WFS-session --type session --content '{"status":"active","resumed_at":"2025-12-10T08:00:00Z"}'
 ```
 
-### Step 5: Add Resume Timestamp
-```bash
-jq '.resumed_at = "'$(date -u +%Y-%m-%dT%H:%M:%SZ)'"' .workflow/active/WFS-session/workflow-session.json > temp.json
-mv temp.json .workflow/active/WFS-session/workflow-session.json
-```
-
-## Simple Bash Commands
+## Simple Commands
 
 ### Basic Operations
-- **List sessions**: `ls .workflow/active/WFS-*`
-- **Check status**: `jq -r '.status' session.json`
-- **Find recent**: `ls -t .workflow/active/*/workflow-session.json | head -1`
-- **Update status**: `jq '.status = "active"' session.json > temp.json`
-- **Add timestamp**: `jq '.resumed_at = "'$(date -u +%Y-%m-%dT%H:%M:%SZ)'"'`
+- **List sessions**: `ccw session list --location active`
+- **Check status**: `ccw session read WFS-xxx --type session`
+- **Update status**: `ccw session status WFS-xxx active`
 
 ### Resume Result
 ```
@@ -58,4 +54,26 @@ Session WFS-user-auth resumed
 - Paused at: 2025-09-15T14:30:00Z
 - Resumed at: 2025-09-15T15:45:00Z
 - Ready for: /workflow:execute
-```
\ No newline at end of file
+```
+## session_manager Tool Alternative
+
+Use `ccw tool exec session_manager` for session resume:
+
+### Update Session Status
+```bash
+# Update status to active
+ccw tool exec session_manager '{
+  "operation": "update",
+  "session_id": "WFS-xxx",
+  "content_type": "session",
+  "content": {
+    "status": "active",
+    "resumed_at": "2025-12-10T08:00:00Z"
+  }
+}'
+```
+
+### Read Session Status
+```bash
+ccw tool exec session_manager '{"operation":"read","session_id":"WFS-xxx","content_type":"session"}'
+```
diff --git a/.claude/skills/command-guide/reference/commands/workflow/session/start.md b/.claude/skills/command-guide/reference/commands/workflow/session/start.md
index 7837d6f0..ec053208 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/session/start.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/session/start.md
@@ -70,12 +70,12 @@ SlashCommand({command: "/workflow:init"});
 
 ### Step 1: List Active Sessions
 ```bash
-bash(ls -1 .workflow/active/ 2>/dev/null | head -5)
+ccw session list --location active
 ```
 
 ### Step 2: Display Session Metadata
 ```bash
-bash(cat .workflow/active/WFS-promptmaster-platform/workflow-session.json)
+ccw session read WFS-promptmaster-platform --type session
 ```
 
 ### Step 4: User Decision
@@ -92,34 +92,29 @@ Present session information and wait for user to select or create session.
 
 ### Step 1: Check Active Sessions Count
 ```bash
-bash(find .workflow/active/ -name "WFS-*" -type d 2>/dev/null | wc -l)
+ccw session list --location active
+# Check result.total in response
 ```
 
 ### Step 2a: No Active Sessions → Create New
 ```bash
-# Generate session slug
-bash(echo "implement OAuth2 auth" | sed 's/[^a-zA-Z0-9]/-/g' | tr '[:upper:]' '[:lower:]' | cut -c1-50)
+# Generate session slug from description
+# Pattern: WFS-{lowercase-slug-from-description}
 
-# Create directory structure
-bash(mkdir -p .workflow/active/WFS-implement-oauth2-auth/.process)
-bash(mkdir -p .workflow/active/WFS-implement-oauth2-auth/.task)
-bash(mkdir -p .workflow/active/WFS-implement-oauth2-auth/.summaries)
-
-# Create metadata (include type field, default to "workflow" if not specified)
-bash(echo '{"session_id":"WFS-implement-oauth2-auth","project":"implement OAuth2 auth","status":"planning","type":"workflow","created_at":"2024-12-04T08:00:00Z"}' > .workflow/active/WFS-implement-oauth2-auth/workflow-session.json)
+# Create session with ccw (creates directories + metadata atomically)
+ccw session init WFS-implement-oauth2-auth --type workflow --content '{"project":"implement OAuth2 auth","status":"planning"}'
 ```
 
 **Output**: `SESSION_ID: WFS-implement-oauth2-auth`
 
 ### Step 2b: Single Active Session → Check Relevance
 ```bash
-# Extract session ID
-bash(find .workflow/active/ -name "WFS-*" -type d 2>/dev/null | head -1 | xargs basename)
+# Get session list with metadata
+ccw session list --location active
 
-# Read project name from metadata
-bash(cat .workflow/active/WFS-promptmaster-platform/workflow-session.json | grep -o '"project":"[^"]*"' | cut -d'"' -f4)
+# Read session metadata for relevance check
+ccw session read WFS-promptmaster-platform --type session
 
-# Check keyword match (manual comparison)
 # If task contains project keywords → Reuse session
 # If task unrelated → Create new session (use Step 2a)
 ```
@@ -129,8 +124,9 @@ bash(cat .workflow/active/WFS-promptmaster-platform/workflow-session.json | grep
 
 ### Step 2c: Multiple Active Sessions → Use First
 ```bash
-# Get first active session
-bash(find .workflow/active/ -name "WFS-*" -type d 2>/dev/null | head -1 | xargs basename)
+# Get first active session from list
+ccw session list --location active
+# Use first session_id from result.active array
 
 # Output warning and session ID
 # WARNING: Multiple active sessions detected
@@ -146,24 +142,15 @@ bash(find .workflow/active/ -name "WFS-*" -type d 2>/dev/null | head -1 | xargs
 
 ### Step 1: Generate Unique Session Slug
 ```bash
-# Convert to slug
-bash(echo "fix login bug" | sed 's/[^a-zA-Z0-9]/-/g' | tr '[:upper:]' '[:lower:]' | cut -c1-50)
-
-# Check if exists, add counter if needed
-bash(ls .workflow/active/WFS-fix-login-bug 2>/dev/null && echo "WFS-fix-login-bug-2" || echo "WFS-fix-login-bug")
+# Convert description to slug: lowercase, alphanumeric + hyphen, max 50 chars
+# Check if exists via ccw session list, add counter if collision
+ccw session list --location active
 ```
 
 ### Step 2: Create Session Structure
 ```bash
-bash(mkdir -p .workflow/active/WFS-fix-login-bug/.process)
-bash(mkdir -p .workflow/active/WFS-fix-login-bug/.task)
-bash(mkdir -p .workflow/active/WFS-fix-login-bug/.summaries)
-```
-
-### Step 3: Create Metadata
-```bash
-# Include type field from --type parameter (default: "workflow")
-bash(echo '{"session_id":"WFS-fix-login-bug","project":"fix login bug","status":"planning","type":"workflow","created_at":"2024-12-04T08:00:00Z"}' > .workflow/active/WFS-fix-login-bug/workflow-session.json)
+# Single command creates directories (.process, .task, .summaries) + metadata
+ccw session init WFS-fix-login-bug --type workflow --content '{"project":"fix login bug","status":"planning"}'
 ```
 
 **Output**: `SESSION_ID: WFS-fix-login-bug`
@@ -197,4 +184,36 @@ SESSION_ID: WFS-promptmaster-platform
 - Pattern: `WFS-[lowercase-slug]`
 - Characters: `a-z`, `0-9`, `-` only
 - Max length: 50 characters
-- Uniqueness: Add numeric suffix if collision (`WFS-auth-2`, `WFS-auth-3`)
\ No newline at end of file
+- Uniqueness: Add numeric suffix if collision (`WFS-auth-2`, `WFS-auth-3`)
+
+## session_manager Tool Alternative
+
+The above bash commands can be replaced with `ccw tool exec session_manager`:
+
+### List Sessions
+```bash
+# List active sessions with metadata
+ccw tool exec session_manager '{"operation":"list","location":"active","include_metadata":true}'
+
+# Response: {"success":true,"result":{"active":[{"session_id":"WFS-xxx","metadata":{...}}],"total":1}}
+```
+
+### Create Session (replaces mkdir + echo)
+```bash
+# Single command creates directories + metadata
+ccw tool exec session_manager '{
+  "operation": "init",
+  "session_id": "WFS-my-session",
+  "metadata": {
+    "project": "my project description",
+    "status": "planning",
+    "type": "workflow",
+    "created_at": "2025-12-10T08:00:00Z"
+  }
+}'
+```
+
+### Read Session Metadata
+```bash
+ccw tool exec session_manager '{"operation":"read","session_id":"WFS-xxx","content_type":"session"}'
+```
diff --git a/.claude/skills/command-guide/reference/commands/workflow/tdd-plan.md b/.claude/skills/command-guide/reference/commands/workflow/tdd-plan.md
index c913401f..d0e7ff90 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/tdd-plan.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/tdd-plan.md
@@ -164,10 +164,10 @@ SlashCommand(command="/workflow:tools:conflict-resolution --session [sessionId]
 
 **Parse Output**:
 - Extract: Execution status (success/skipped/failed)
-- Verify: CONFLICT_RESOLUTION.md file path (if executed)
+- Verify: conflict-resolution.json file path (if executed)
 
 **Validation**:
-- File `.workflow/active/[sessionId]/.process/CONFLICT_RESOLUTION.md` exists (if executed)
+- File `.workflow/active/[sessionId]/.process/conflict-resolution.json` exists (if executed)
 
 **Skip Behavior**:
 - If conflict_risk is "none" or "low", skip directly to Phase 5
@@ -402,7 +402,7 @@ TDD Workflow Orchestrator
 │     ├─ Phase 4.1: Detect conflicts with CLI
 │     ├─ Phase 4.2: Present conflicts to user
 │     └─ Phase 4.3: Apply resolution strategies
-│     └─ Returns: CONFLICT_RESOLUTION.md              ← COLLAPSED
+│     └─ Returns: conflict-resolution.json            ← COLLAPSED
 │  ELSE:
 │  └─ Skip to Phase 5
 │
diff --git a/.claude/skills/command-guide/reference/commands/workflow/tdd-verify.md b/.claude/skills/command-guide/reference/commands/workflow/tdd-verify.md
index 587ee3f9..6c51e07c 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/tdd-verify.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/tdd-verify.md
@@ -127,7 +127,7 @@ ccw session read {sessionId} --type task --raw | jq -r '.meta.agent'
 **Gemini analysis for comprehensive TDD compliance report**
 
 ```bash
-cd project-root && gemini -p "
+ccw cli exec "
 PURPOSE: Generate TDD compliance report
 TASK: Analyze TDD workflow execution and generate quality report
 CONTEXT: @{.workflow/active/{sessionId}/.task/*.json,.workflow/active/{sessionId}/.summaries/*,.workflow/active/{sessionId}/.process/tdd-cycle-report.md}
@@ -139,7 +139,7 @@ EXPECTED:
 - Red-Green-Refactor cycle validation
 - Best practices adherence assessment
 RULES: Focus on TDD best practices and workflow adherence. Be specific about violations and improvements.
-" > .workflow/active/{sessionId}/TDD_COMPLIANCE_REPORT.md
+" --tool gemini --cd project-root > .workflow/active/{sessionId}/TDD_COMPLIANCE_REPORT.md
 ```
 
 **Output**: TDD_COMPLIANCE_REPORT.md
diff --git a/.claude/skills/command-guide/reference/commands/workflow/tools/conflict-resolution.md b/.claude/skills/command-guide/reference/commands/workflow/tools/conflict-resolution.md
index 85650e86..dc6daa16 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/tools/conflict-resolution.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/tools/conflict-resolution.md
@@ -133,7 +133,7 @@ Task(subagent_type="cli-execution-agent", prompt=`
   ### 2. Execute CLI Analysis (Enhanced with Exploration + Scenario Uniqueness)
 
   Primary (Gemini):
-  cd {project_root} && gemini -p "
+  ccw cli exec "
   PURPOSE: Detect conflicts between plan and codebase, using exploration insights
   TASK:
   • **Review pre-identified conflict_indicators from exploration results**
@@ -152,7 +152,7 @@ Task(subagent_type="cli-execution-agent", prompt=`
     - ModuleOverlap conflicts with overlap_analysis
     - Targeted clarification questions
   RULES: $(cat ~/.claude/workflows/cli-templates/prompts/analysis/02-analyze-code-patterns.txt) | Focus on breaking changes, migration needs, and functional overlaps | Prioritize exploration-identified conflicts | analysis=READ-ONLY
-  "
+  " --tool gemini --cd {project_root}
 
   Fallback: Qwen (same prompt) → Claude (manual analysis)
 
@@ -169,7 +169,7 @@ Task(subagent_type="cli-execution-agent", prompt=`
 
   ### 4. Return Structured Conflict Data
 
-  ⚠️ DO NOT generate CONFLICT_RESOLUTION.md file
+  ⚠️ Output to conflict-resolution.json (generated in Phase 4)
 
   Return JSON format for programmatic processing:
 
@@ -467,14 +467,30 @@ selectedStrategies.forEach(item => {
 
 console.log(`\n正在应用 ${modifications.length} 个修改...`);
 
-// 2. Apply each modification using Edit tool
+// 2. Apply each modification using Edit tool (with fallback to context-package.json)
 const appliedModifications = [];
 const failedModifications = [];
+const fallbackConstraints = [];  // For files that don't exist
 
 modifications.forEach((mod, idx) => {
   try {
     console.log(`[${idx + 1}/${modifications.length}] 修改 ${mod.file}...`);
 
+    // Check if target file exists (brainstorm files may not exist in lite workflow)
+    if (!file_exists(mod.file)) {
+      console.log(`  ⚠️ 文件不存在，写入 context-package.json 作为约束`);
+      fallbackConstraints.push({
+        source: "conflict-resolution",
+        conflict_id: mod.conflict_id,
+        target_file: mod.file,
+        section: mod.section,
+        change_type: mod.change_type,
+        content: mod.new_content,
+        rationale: mod.rationale
+      });
+      return;  // Skip to next modification
+    }
+
     if (mod.change_type === "update") {
       Edit({
         file_path: mod.file,
@@ -502,14 +518,45 @@ modifications.forEach((mod, idx) => {
   }
 });
 
-// 3. Update context-package.json with resolution details
+// 2b. Generate conflict-resolution.json output file
+const resolutionOutput = {
+  session_id: sessionId,
+  resolved_at: new Date().toISOString(),
+  summary: {
+    total_conflicts: conflicts.length,
+    resolved_with_strategy: selectedStrategies.length,
+    custom_handling: customConflicts.length,
+    fallback_constraints: fallbackConstraints.length
+  },
+  resolved_conflicts: selectedStrategies.map(s => ({
+    conflict_id: s.conflict_id,
+    strategy_name: s.strategy.name,
+    strategy_approach: s.strategy.approach,
+    clarifications: s.clarifications || [],
+    modifications_applied: s.strategy.modifications?.filter(m =>
+      appliedModifications.some(am => am.conflict_id === s.conflict_id)
+    ) || []
+  })),
+  custom_conflicts: customConflicts.map(c => ({
+    id: c.id,
+    brief: c.brief,
+    category: c.category,
+    suggestions: c.suggestions,
+    overlap_analysis: c.overlap_analysis || null
+  })),
+  planning_constraints: fallbackConstraints,  // Constraints for files that don't exist
+  failed_modifications: failedModifications
+};
+
+const resolutionPath = `.workflow/active/${sessionId}/.process/conflict-resolution.json`;
+Write(resolutionPath, JSON.stringify(resolutionOutput, null, 2));
+console.log(`\n📄 冲突解决结果已保存: ${resolutionPath}`);
+
+// 3. Update context-package.json with resolution details (reference to JSON file)
 const contextPackage = JSON.parse(Read(contextPath));
 contextPackage.conflict_detection.conflict_risk = "resolved";
-contextPackage.conflict_detection.resolved_conflicts = selectedStrategies.map(s => ({
-  conflict_id: s.conflict_id,
-  strategy_name: s.strategy.name,
-  clarifications: s.clarifications
-}));
+contextPackage.conflict_detection.resolution_file = resolutionPath;  // Reference to detailed JSON
+contextPackage.conflict_detection.resolved_conflicts = selectedStrategies.map(s => s.conflict_id);
 contextPackage.conflict_detection.custom_conflicts = customConflicts.map(c => c.id);
 contextPackage.conflict_detection.resolved_at = new Date().toISOString();
 Write(contextPath, JSON.stringify(contextPackage, null, 2));
@@ -582,12 +629,50 @@ return {
 ✓ Agent log saved to .workflow/active/{session_id}/.chat/
 ```
 
-## Output Format: Agent JSON Response
+## Output Format
+
+### Primary Output: conflict-resolution.json
+
+**Path**: `.workflow/active/{session_id}/.process/conflict-resolution.json`
+
+**Schema**:
+```json
+{
+  "session_id": "WFS-xxx",
+  "resolved_at": "ISO timestamp",
+  "summary": {
+    "total_conflicts": 3,
+    "resolved_with_strategy": 2,
+    "custom_handling": 1,
+    "fallback_constraints": 0
+  },
+  "resolved_conflicts": [
+    {
+      "conflict_id": "CON-001",
+      "strategy_name": "策略名称",
+      "strategy_approach": "实现方法",
+      "clarifications": [],
+      "modifications_applied": []
+    }
+  ],
+  "custom_conflicts": [
+    {
+      "id": "CON-002",
+      "brief": "冲突摘要",
+      "category": "ModuleOverlap",
+      "suggestions": ["建议1", "建议2"],
+      "overlap_analysis": null
+    }
+  ],
+  "planning_constraints": [],
+  "failed_modifications": []
+}
+```
+
+### Secondary: Agent JSON Response (stdout)
 
 **Focus**: Structured conflict data with actionable modifications for programmatic processing.
 
-**Format**: JSON to stdout (NO file generation)
-
 **Structure**: Defined in Phase 2, Step 4 (agent prompt)
 
 ### Key Requirements
@@ -635,11 +720,12 @@ If Edit tool fails mid-application:
 - Requires: `conflict_risk ≥ medium`
 
 **Output**:
-- Modified files:
+- Generated file:
+  - `.workflow/active/{session_id}/.process/conflict-resolution.json` (primary output)
+- Modified files (if exist):
   - `.workflow/active/{session_id}/.brainstorm/guidance-specification.md`
   - `.workflow/active/{session_id}/.brainstorm/{role}/analysis.md`
-  - `.workflow/active/{session_id}/.process/context-package.json` (conflict_risk → resolved)
-- NO report file generation
+  - `.workflow/active/{session_id}/.process/context-package.json` (conflict_risk → resolved, resolution_file reference)
 
 **User Interaction**:
 - **Iterative conflict processing**: One conflict at a time, not in batches
@@ -667,7 +753,7 @@ If Edit tool fails mid-application:
 ✓ guidance-specification.md updated with resolved conflicts
 ✓ Role analyses (*.md) updated with resolved conflicts
 ✓ context-package.json marked as "resolved" with clarification records
-✓ No CONFLICT_RESOLUTION.md file generated
+✓ conflict-resolution.json generated with full resolution details
 ✓ Modification summary includes:
   - Total conflicts
   - Resolved with strategy (count)
diff --git a/.claude/skills/command-guide/reference/commands/workflow/tools/task-generate-agent.md b/.claude/skills/command-guide/reference/commands/workflow/tools/task-generate-agent.md
index d8ad0cad..16f72c43 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/tools/task-generate-agent.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/tools/task-generate-agent.md
@@ -89,6 +89,14 @@ Phase 3: Integration (+1 Coordinator, Multi-Module Only)
 3. **Auto Module Detection** (determines single vs parallel mode):
    ```javascript
    function autoDetectModules(contextPackage, projectRoot) {
+     // === Complexity Gate: Only parallelize for High complexity ===
+     const complexity = contextPackage.metadata?.complexity || 'Medium';
+     if (complexity !== 'High') {
+       // Force single agent mode for Low/Medium complexity
+       // This maximizes agent context reuse for related tasks
+       return [{ name: 'main', prefix: '', paths: ['.'] }];
+     }
+
      // Priority 1: Explicit frontend/backend separation
      if (exists('src/frontend') && exists('src/backend')) {
        return [
@@ -112,8 +120,9 @@ Phase 3: Integration (+1 Coordinator, Multi-Module Only)
    ```
 
 **Decision Logic**:
+- `complexity !== 'High'` → Force Phase 2A (Single Agent, maximize context reuse)
 - `modules.length == 1` → Phase 2A (Single Agent, original flow)
-- `modules.length >= 2` → Phase 2B + Phase 3 (N+1 Parallel)
+- `modules.length >= 2 && complexity == 'High'` → Phase 2B + Phase 3 (N+1 Parallel)
 
 **Note**: CLI tool usage is now determined semantically by action-planning-agent based on user's task description, not by flags.
 
@@ -163,6 +172,13 @@ Determine CLI tool usage per-step based on user's task description:
 - Use aggregated_insights.all_integration_points for precise modification locations
 - Use conflict_indicators for risk-aware task sequencing
 
+## CONFLICT RESOLUTION CONTEXT (if exists)
+- Check context-package.conflict_detection.resolution_file for conflict-resolution.json path
+- If exists, load .process/conflict-resolution.json:
+  - Apply planning_constraints as task constraints (for brainstorm-less workflows)
+  - Reference resolved_conflicts for implementation approach alignment
+  - Handle custom_conflicts with explicit task notes
+
 ## EXPECTED DELIVERABLES
 1. Task JSON Files (.task/IMPL-*.json)
    - 6-field schema (id, title, status, context_package_path, meta, context, flow_control)
@@ -203,7 +219,9 @@ Hard Constraints:
 
 **Condition**: `modules.length >= 2` (multi-module detected)
 
-**Purpose**: Launch N action-planning-agents simultaneously, one per module, for parallel task generation.
+**Purpose**: Launch N action-planning-agents simultaneously, one per module, for parallel task JSON generation.
+
+**Note**: Phase 2B agents generate Task JSONs ONLY. IMPL_PLAN.md and TODO_LIST.md are generated by Phase 3 Coordinator.
 
 **Parallel Agent Invocation**:
 ```javascript
@@ -211,27 +229,49 @@ Hard Constraints:
 const planningTasks = modules.map(module =>
   Task(
     subagent_type="action-planning-agent",
-    description=`Plan ${module.name} module`,
+    description=`Generate ${module.name} module task JSONs`,
     prompt=`
-## SCOPE
+## TASK OBJECTIVE
+Generate task JSON files for ${module.name} module within workflow session
+
+IMPORTANT: This is PLANNING ONLY - generate task JSONs, NOT implementing code.
+IMPORTANT: Generate Task JSONs ONLY. IMPL_PLAN.md and TODO_LIST.md by Phase 3 Coordinator.
+
+CRITICAL: Follow progressive loading strategy in agent specification
+
+## MODULE SCOPE
 - Module: ${module.name} (${module.type})
 - Focus Paths: ${module.paths.join(', ')}
 - Task ID Prefix: IMPL-${module.prefix}
 - Task Limit: ≤9 tasks
 - Other Modules: ${otherModules.join(', ')}
-- Cross-module deps format: "CROSS::{module}::{pattern}"
 
 ## SESSION PATHS
 Input:
+  - Session Metadata: .workflow/active/{session-id}/workflow-session.json
   - Context Package: .workflow/active/{session-id}/.process/context-package.json
 Output:
   - Task Dir: .workflow/active/{session-id}/.task/
 
-## INSTRUCTIONS
-- Generate tasks ONLY for ${module.name} module
-- Use task ID format: IMPL-${module.prefix}1, IMPL-${module.prefix}2, ...
-- For cross-module dependencies, use: depends_on: ["CROSS::B::api-endpoint"]
-- Maximum 9 tasks per module
+## CONTEXT METADATA
+Session ID: {session-id}
+MCP Capabilities: {exa_code, exa_web, code_index}
+
+## CROSS-MODULE DEPENDENCIES
+- Use placeholder: depends_on: ["CROSS::{module}::{pattern}"]
+- Example: depends_on: ["CROSS::B::api-endpoint"]
+- Phase 3 Coordinator resolves to actual task IDs
+
+## EXPECTED DELIVERABLES
+Task JSON Files (.task/IMPL-${module.prefix}*.json):
+  - 6-field schema per agent specification
+  - Task ID format: IMPL-${module.prefix}1, IMPL-${module.prefix}2, ...
+  - Focus ONLY on ${module.name} module scope
+
+## SUCCESS CRITERIA
+- Task JSONs saved to .task/ with IMPL-${module.prefix}* naming
+- Cross-module dependencies use CROSS:: placeholder format
+- Return task count and brief summary
     `
   )
 );
@@ -255,37 +295,78 @@ await Promise.all(planningTasks);
 - Prefix: A, B, C... (assigned by detection order)
 - Sequence: 1, 2, 3... (per-module increment)
 
-### Phase 3: Integration (+1 Coordinator, Multi-Module Only)
+### Phase 3: Integration (+1 Coordinator Agent, Multi-Module Only)
 
 **Condition**: Only executed when `modules.length >= 2`
 
-**Purpose**: Collect all module tasks, resolve cross-module dependencies, generate unified documents.
+**Purpose**: Collect all module tasks, resolve cross-module dependencies, generate unified IMPL_PLAN.md and TODO_LIST.md documents.
 
-**Integration Logic**:
+**Coordinator Agent Invocation**:
 ```javascript
-// 1. Collect all module task JSONs
-const allTasks = glob('.task/IMPL-*.json').map(loadJson);
+// Wait for all Phase 2B agents to complete
+const moduleResults = await Promise.all(planningTasks);
 
-// 2. Resolve cross-module dependencies
-for (const task of allTasks) {
-  if (task.depends_on) {
-    task.depends_on = task.depends_on.map(dep => {
-      if (dep.startsWith('CROSS::')) {
-        // CROSS::B::api-endpoint → find matching IMPL-B* task
-        const [, targetModule, pattern] = dep.match(/CROSS::(\w+)::(.+)/);
-        return findTaskByModuleAndPattern(allTasks, targetModule, pattern);
-      }
-      return dep;
-    });
-  }
-}
+// Launch +1 Coordinator Agent
+Task(
+  subagent_type="action-planning-agent",
+  description="Integrate module tasks and generate unified documents",
+  prompt=`
+## TASK OBJECTIVE
+Integrate all module task JSONs, resolve cross-module dependencies, and generate unified IMPL_PLAN.md and TODO_LIST.md
 
-// 3. Generate unified IMPL_PLAN.md (grouped by module)
-generateIMPL_PLAN(allTasks, modules);
+IMPORTANT: This is INTEGRATION ONLY - consolidate existing task JSONs, NOT creating new tasks.
 
-// 4. Generate TODO_LIST.md (hierarchical structure)
-generateTODO_LIST(allTasks, modules);
+## SESSION PATHS
+Input:
+  - Session Metadata: .workflow/active/{session-id}/workflow-session.json
+  - Context Package: .workflow/active/{session-id}/.process/context-package.json
+  - Task JSONs: .workflow/active/{session-id}/.task/IMPL-*.json (from Phase 2B)
+Output:
+  - Updated Task JSONs: .workflow/active/{session-id}/.task/IMPL-*.json (resolved dependencies)
+  - IMPL_PLAN: .workflow/active/{session-id}/IMPL_PLAN.md
+  - TODO_LIST: .workflow/active/{session-id}/TODO_LIST.md
+
+## CONTEXT METADATA
+Session ID: {session-id}
+Modules: ${modules.map(m => m.name + '(' + m.prefix + ')').join(', ')}
+Module Count: ${modules.length}
+
+## INTEGRATION STEPS
+1. Collect all .task/IMPL-*.json, group by module prefix
+2. Resolve CROSS:: dependencies → actual task IDs, update task JSONs
+3. Generate IMPL_PLAN.md (multi-module format per agent specification)
+4. Generate TODO_LIST.md (hierarchical format per agent specification)
+
+## CROSS-MODULE DEPENDENCY RESOLUTION
+- Pattern: CROSS::{module}::{pattern} → IMPL-{module}* matching title/context
+- Example: CROSS::B::api-endpoint → IMPL-B1 (if B1 title contains "api-endpoint")
+- Log unresolved as warnings
+
+## EXPECTED DELIVERABLES
+1. Updated Task JSONs with resolved dependency IDs
+2. IMPL_PLAN.md - multi-module format with cross-dependency section
+3. TODO_LIST.md - hierarchical by module with cross-dependency section
+
+## SUCCESS CRITERIA
+- No CROSS:: placeholders remaining in task JSONs
+- IMPL_PLAN.md and TODO_LIST.md generated with multi-module structure
+- Return: task count, per-module breakdown, resolved dependency count
+  `
+)
 ```
 
-**Note**: IMPL_PLAN.md and TODO_LIST.md structure definitions are in `action-planning-agent.md`.
+**Dependency Resolution Algorithm**:
+```javascript
+function resolveCrossModuleDependency(placeholder, allTasks) {
+  const [, targetModule, pattern] = placeholder.match(/CROSS::(\w+)::(.+)/);
+  const candidates = allTasks.filter(t =>
+    t.id.startsWith(`IMPL-${targetModule}`) &&
+    (t.title.toLowerCase().includes(pattern.toLowerCase()) ||
+     t.context?.description?.toLowerCase().includes(pattern.toLowerCase()))
+  );
+  return candidates.length > 0
+    ? candidates.sort((a, b) => a.id.localeCompare(b.id))[0].id
+    : placeholder; // Keep for manual resolution
+}
+```
 
diff --git a/.claude/skills/command-guide/reference/commands/workflow/tools/task-generate-tdd.md b/.claude/skills/command-guide/reference/commands/workflow/tools/task-generate-tdd.md
index e1fe947c..177eae6c 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/tools/task-generate-tdd.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/tools/task-generate-tdd.md
@@ -152,9 +152,14 @@ Phase 2: Agent Execution (Document Generation)
    roleAnalysisPaths.forEach(path => Read(path));
    ```
 
-5. **Load Conflict Resolution** (from context-package.json, if exists)
+5. **Load Conflict Resolution** (from conflict-resolution.json, if exists)
    ```javascript
-   if (contextPackage.brainstorm_artifacts.conflict_resolution?.exists) {
+   // Check for new conflict-resolution.json format
+   if (contextPackage.conflict_detection?.resolution_file) {
+     Read(contextPackage.conflict_detection.resolution_file)  // .process/conflict-resolution.json
+   }
+   // Fallback: legacy brainstorm_artifacts path
+   else if (contextPackage.brainstorm_artifacts?.conflict_resolution?.exists) {
      Read(contextPackage.brainstorm_artifacts.conflict_resolution.path)
    }
    ```
@@ -223,7 +228,7 @@ If conflict_risk was medium/high, modifications have been applied to:
 - **guidance-specification.md**: Design decisions updated to resolve conflicts
 - **Role analyses (*.md)**: Recommendations adjusted for compatibility
 - **context-package.json**: Marked as "resolved" with conflict IDs
-- NO separate CONFLICT_RESOLUTION.md file (conflicts resolved in-place)
+- Conflict resolution results stored in conflict-resolution.json
 
 ### MCP Analysis Results (Optional)
 **Code Structure**: {mcp_code_index_results}
@@ -373,10 +378,12 @@ const agentContext = {
     .flatMap(role => role.files)
     .map(file => Read(file.path)),
 
-  // Load conflict resolution if exists (from context package)
-  conflict_resolution: brainstorm_artifacts.conflict_resolution?.exists
-    ? Read(brainstorm_artifacts.conflict_resolution.path)
-    : null,
+  // Load conflict resolution if exists (prefer new JSON format)
+  conflict_resolution: context_package.conflict_detection?.resolution_file
+    ? Read(context_package.conflict_detection.resolution_file)  // .process/conflict-resolution.json
+    : (brainstorm_artifacts?.conflict_resolution?.exists
+        ? Read(brainstorm_artifacts.conflict_resolution.path)
+        : null),
 
   // Optional MCP enhancements
   mcp_analysis: executeMcpDiscovery()
@@ -408,7 +415,7 @@ This section provides quick reference for TDD task JSON structure. For complete
 │   ├── IMPL-3.2.json                # Complex feature subtask (if needed)
 │   └── ...
 └── .process/
-    ├── CONFLICT_RESOLUTION.md       # Conflict resolution strategies (if conflict_risk ≥ medium)
+    ├── conflict-resolution.json     # Conflict resolution results (if conflict_risk ≥ medium)
     ├── test-context-package.json    # Test coverage analysis
     ├── context-package.json         # Input from context-gather
     ├── context_package_path         # Path to smart context package
diff --git a/.claude/skills/command-guide/reference/commands/workflow/tools/test-concept-enhanced.md b/.claude/skills/command-guide/reference/commands/workflow/tools/test-concept-enhanced.md
index 74a5b786..74338925 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/tools/test-concept-enhanced.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/tools/test-concept-enhanced.md
@@ -89,7 +89,7 @@ Template: ~/.claude/workflows/cli-templates/prompts/test/test-concept-analysis.t
 
 ## EXECUTION STEPS
 1. Execute Gemini analysis:
-   cd .workflow/active/{test_session_id}/.process && gemini -p "$(cat ~/.claude/workflows/cli-templates/prompts/test/test-concept-analysis.txt)" --approval-mode yolo
+   ccw cli exec "$(cat ~/.claude/workflows/cli-templates/prompts/test/test-concept-analysis.txt)" --tool gemini --mode write --cd .workflow/active/{test_session_id}/.process
 
 2. Generate TEST_ANALYSIS_RESULTS.md:
    Synthesize gemini-test-analysis.md into standardized format for task generation
diff --git a/.claude/skills/command-guide/reference/commands/workflow/ui-design/import-from-code.md b/.claude/skills/command-guide/reference/commands/workflow/ui-design/import-from-code.md
index 71e42bcc..9b86f03e 100644
--- a/.claude/skills/command-guide/reference/commands/workflow/ui-design/import-from-code.md
+++ b/.claude/skills/command-guide/reference/commands/workflow/ui-design/import-from-code.md
@@ -180,14 +180,14 @@ Task(subagent_type="ui-design-agent",
   - Pattern: rg → Extract values → Compare → If different → Read full context with comments → Record conflict
   - Alternative (if many files): Execute CLI analysis for comprehensive report:
     \`\`\`bash
-    cd ${source} && gemini -p \"
+    ccw cli exec \"
     PURPOSE: Detect color token conflicts across all CSS/SCSS/JS files
     TASK: • Scan all files for color definitions • Identify conflicting values • Extract semantic comments
     MODE: analysis
     CONTEXT: @**/*.css @**/*.scss @**/*.js @**/*.ts
     EXPECTED: JSON report listing conflicts with file:line, values, semantic context
     RULES: Focus on core tokens | Report ALL variants | analysis=READ-ONLY
-    \"
+    \" --tool gemini --cd ${source}
     \`\`\`
 
   **Step 1: Load file list**
@@ -295,14 +295,14 @@ Task(subagent_type="ui-design-agent",
   - Pattern: rg → Identify animation types → Map framework usage → Prioritize extraction targets
   - Alternative (if complex framework mix): Execute CLI analysis for comprehensive report:
     \`\`\`bash
-    cd ${source} && gemini -p \"
+    ccw cli exec \"
     PURPOSE: Detect animation frameworks and patterns
     TASK: • Identify frameworks • Map animation patterns • Categorize by complexity
     MODE: analysis
     CONTEXT: @**/*.css @**/*.scss @**/*.js @**/*.ts
     EXPECTED: JSON report listing frameworks, animation types, file locations
     RULES: Focus on framework consistency | Map all animations | analysis=READ-ONLY
-    \"
+    \" --tool gemini --cd ${source}
     \`\`\`
 
   **Step 1: Load file list**
@@ -374,14 +374,14 @@ Task(subagent_type="ui-design-agent",
   - Pattern: rg → Count occurrences → Classify by frequency → Prioritize universal components
   - Alternative (if large codebase): Execute CLI analysis for comprehensive categorization:
     \`\`\`bash
-    cd ${source} && gemini -p \"
+    ccw cli exec \"
     PURPOSE: Classify components as universal vs specialized
     TASK: • Identify UI components • Classify reusability • Map layout systems
     MODE: analysis
     CONTEXT: @**/*.css @**/*.scss @**/*.js @**/*.ts @**/*.html
     EXPECTED: JSON report categorizing components, layout patterns, naming conventions
     RULES: Focus on component reusability | Identify layout systems | analysis=READ-ONLY
-    \"
+    \" --tool gemini --cd ${source}
     \`\`\`
 
   **Step 1: Load file list**
diff --git a/ccw/src/core/dashboard-generator.ts b/ccw/src/core/dashboard-generator.ts
index ef56e6ce..c0641285 100644
--- a/ccw/src/core/dashboard-generator.ts
+++ b/ccw/src/core/dashboard-generator.ts
@@ -6,12 +6,12 @@ import { fileURLToPath } from 'url';
 const __filename = fileURLToPath(import.meta.url);
 const __dirname = dirname(__filename);
 
-// Bundled template paths
-const UNIFIED_TEMPLATE = join(__dirname, '../templates/dashboard.html');
-const JS_FILE = join(__dirname, '../templates/dashboard.js');
-const MODULE_CSS_DIR = join(__dirname, '../templates/dashboard-css');
-const WORKFLOW_TEMPLATE = join(__dirname, '../templates/workflow-dashboard.html');
-const REVIEW_TEMPLATE = join(__dirname, '../templates/review-cycle-dashboard.html');
+// Bundled template paths (from dist/core/ -> src/templates/)
+const UNIFIED_TEMPLATE = join(__dirname, '../../src/templates/dashboard.html');
+const JS_FILE = join(__dirname, '../../src/templates/dashboard.js');
+const MODULE_CSS_DIR = join(__dirname, '../../src/templates/dashboard-css');
+const WORKFLOW_TEMPLATE = join(__dirname, '../../src/templates/workflow-dashboard.html');
+const REVIEW_TEMPLATE = join(__dirname, '../../src/templates/review-cycle-dashboard.html');
 
 // Modular CSS files in load order
 const MODULE_CSS_FILES = [
diff --git a/ccw/src/core/server.ts b/ccw/src/core/server.ts
index 754d8ca1..baceda85 100644
--- a/ccw/src/core/server.ts
+++ b/ccw/src/core/server.ts
@@ -36,11 +36,11 @@ function getEnterpriseMcpPath(): string {
 // WebSocket clients for real-time notifications
 const wsClients = new Set();
 
-const TEMPLATE_PATH = join(import.meta.dirname, '../templates/dashboard.html');
-const MODULE_CSS_DIR = join(import.meta.dirname, '../templates/dashboard-css');
-const JS_FILE = join(import.meta.dirname, '../templates/dashboard.js');
-const MODULE_JS_DIR = join(import.meta.dirname, '../templates/dashboard-js');
-const ASSETS_DIR = join(import.meta.dirname, '../templates/assets');
+const TEMPLATE_PATH = join(import.meta.dirname, '../../src/templates/dashboard.html');
+const MODULE_CSS_DIR = join(import.meta.dirname, '../../src/templates/dashboard-css');
+const JS_FILE = join(import.meta.dirname, '../../src/templates/dashboard.js');
+const MODULE_JS_DIR = join(import.meta.dirname, '../../src/templates/dashboard-js');
+const ASSETS_DIR = join(import.meta.dirname, '../../src/templates/assets');
 
 // Modular CSS files in load order
 const MODULE_CSS_FILES = [