Refactor CLI Config Manager and Add Provider Model Routes

- Removed deprecated constants and functions from cli-config-manager.ts.
- Introduced new provider model presets in litellm-provider-models.ts for better organization and management of model information.
- Created provider-routes.ts to handle API endpoints for retrieving provider information and models.
- Added integration tests for provider routes to ensure correct functionality and response structure.
- Implemented unit tests for settings persistence functions, covering various scenarios and edge cases.
- Enhanced error handling and validation in the new routes and settings functions.

.claude/agents/code-developer.md

@@ -188,7 +188,7 @@ output               → Variable name to store this step's result
 ```
 // Read task-level execution config (Single Source of Truth)
 const executionMethod = task.meta?.execution_config?.method || 'agent';
-const cliTool = task.meta?.execution_config?.cli_tool || 'codex';
+const cliTool = task.meta?.execution_config?.cli_tool || getDefaultCliTool();  // See ~/.claude/cli-tools.json
 
 // Phase 1: Execute pre_analysis (always by Agent)
 const preAnalysisResults = {};
@@ -240,6 +240,13 @@ ELSE (executionMethod === 'agent'):
 **CLI Handoff Functions**:
 
 ```javascript
+// Get default CLI tool from cli-tools.json
+function getDefaultCliTool() {
+  // Read ~/.claude/cli-tools.json and return first enabled tool
+  // Fallback order: gemini → qwen → codex (first enabled in config)
+  return firstEnabledTool || 'gemini';  // System default fallback
+}
+
 // Build CLI prompt from pre-analysis results and task
 function buildCliHandoffPrompt(preAnalysisResults, task) {
   const contextSection = Object.entries(preAnalysisResults)
@@ -308,7 +315,7 @@ function buildCliCommand(task, cliTool, cliPrompt) {
 | Field | Values | Description |
 |-------|--------|-------------|
 | `method` | `agent` / `cli` / `hybrid` | Execution mode (default: agent) |
-| `cli_tool` | `codex` / `gemini` / `qwen` | CLI tool preference (default: codex) |
+| `cli_tool` | See `~/.claude/cli-tools.json` | CLI tool preference (first enabled tool as default) |
 | `enable_resume` | `true` / `false` | Enable CLI session resume |
 
 **CLI Execution Reference** (from task.cli_execution):
@@ -498,7 +505,8 @@ Before completing any task, verify:
 - Use `run_in_background=false` for all Bash/CLI calls - agent cannot receive task hook callbacks
 - Set timeout ≥60 minutes for CLI commands (hooks don't propagate to subagents):
   ```javascript
-  Bash(command="ccw cli -p '...' --tool codex --mode write", timeout=3600000)  // 60 min
+  Bash(command="ccw cli -p '...' --tool <cli-tool> --mode write", timeout=3600000)  // 60 min
+  // <cli-tool>: First enabled tool from ~/.claude/cli-tools.json (e.g., gemini, qwen, codex)
   ```
 
 **ALWAYS:**

.claude/commands/workflow/execute.md

@@ -477,7 +477,7 @@ Task(subagent_type="{meta.agent}",
      - TODO List: {session.todo_list_path}
      - Summaries: {session.summaries_dir}
 
-     **Execution**: Read task JSON → Parse flow_control → Execute implementation_approach → Update TODO_LIST.md → Generate summary",
+     **Execution**: Read task JSON → Execute pre_analysis → Check execution_config.method → (CLI: handoff to CLI tool | Agent: direct implementation) → Update TODO_LIST.md → Generate summary",
      description="Implement: {task.id}")
 ```
 
@@ -486,9 +486,11 @@ Task(subagent_type="{meta.agent}",
 - `[FLOW_CONTROL]`: Triggers flow_control.pre_analysis execution
 
 **Why Path-Based**: Agent (code-developer.md) autonomously:
-- Reads and parses task JSON (requirements, acceptance, flow_control)
-- Loads tech stack guidelines based on detected language
-- Executes pre_analysis steps and implementation_approach
+- Reads and parses task JSON (requirements, acceptance, flow_control, execution_config)
+- Executes pre_analysis steps (Phase 1: context gathering)
+- Checks execution_config.method (Phase 2: determine mode)
+- CLI mode: Builds handoff prompt and executes via ccw cli with resume strategy
+- Agent mode: Directly implements using modification_points and logic_flow
 - Generates structured summary with integration points
 
 Embedding task content in prompt creates duplication and conflicts with agent's parsing logic.

.claude/commands/workflow/replan.md

@@ -115,14 +115,26 @@ const taskId = taskIdMatch?.[1]
 
 4. **Parse Execution Intent** (from requirements text):
    ```javascript
-   // Extract execution method change from requirements
-   const execPatterns = {
-     cli_codex: /使用\s*(codex|Codex)\s*执行|改用\s*(codex|Codex)/i,
-     cli_gemini: /使用\s*(gemini|Gemini)\s*执行|改用\s*(gemini|Gemini)/i,
-     cli_qwen: /使用\s*(qwen|Qwen)\s*执行|改用\s*(qwen|Qwen)/i,
-     agent: /改为\s*Agent\s*执行|使用\s*Agent\s*执行/i
+   // Dynamic tool detection from cli-tools.json
+   // Read enabled tools: ["gemini", "qwen", "codex", ...]
+   const enabledTools = loadEnabledToolsFromConfig();  // See ~/.claude/cli-tools.json
+
+   // Build dynamic patterns from enabled tools
+   function buildExecPatterns(tools) {
+     const patterns = {
+       agent: /改为\s*Agent\s*执行|使用\s*Agent\s*执行/i
+     };
+     tools.forEach(tool => {
+       // Pattern: "使用 {tool} 执行" or "改用 {tool}"
+       patterns[`cli_${tool}`] = new RegExp(
+         `使用\\s*(${tool})\\s*执行|改用\\s*(${tool})`, 'i'
+       );
+     });
+     return patterns;
    }
 
+   const execPatterns = buildExecPatterns(enabledTools);
+
    let executionIntent = null
    for (const [key, pattern] of Object.entries(execPatterns)) {
      if (pattern.test(requirements)) {