feat: add --mode review support for codex CLI

- Add 'review' to mode enum in ParamsSchema and schema
- Implement codex review subcommand in buildCommand (uses --uncommitted by default)
- Other tools (gemini/qwen/claude) accept review mode but no operation change
- Update cli-tools-usage.md with review mode documentation

ccw/src/commands/cli.ts

@@ -1216,7 +1216,7 @@ export async function cliCommand(
         console.log(chalk.gray('    -f, --file <file>   Read prompt from file (recommended for multi-line prompts)'));
         console.log(chalk.gray('    -p, --prompt <text> Prompt text (single-line)'));
         console.log(chalk.gray('    --tool <tool>       Tool: gemini, qwen, codex (default: gemini)'));
-        console.log(chalk.gray('    --mode <mode>       Mode: analysis, write, auto (default: analysis)'));
+        console.log(chalk.gray('    --mode <mode>       Mode: analysis, write, auto, review (default: analysis)'));
         console.log(chalk.gray('    -d, --debug         Enable debug logging for troubleshooting'));
         console.log(chalk.gray('    --model <model>     Model override'));
         console.log(chalk.gray('    --cd <path>         Working directory'));

ccw/src/tools/cli-executor-core.ts

@@ -160,7 +160,7 @@ interface ClaudeWithSettingsParams {
   prompt: string;
   settingsPath: string;
   endpointId: string;
-  mode: 'analysis' | 'write' | 'auto';
+  mode: 'analysis' | 'write' | 'auto' | 'review';
   workingDir: string;
   cd?: string;
   includeDirs?: string[];
@@ -351,7 +351,7 @@ type BuiltinCliTool = typeof BUILTIN_CLI_TOOLS[number];
 const ParamsSchema = z.object({
   tool: z.string().min(1, 'Tool is required'), // Accept any tool ID (built-in or custom endpoint)
   prompt: z.string().min(1, 'Prompt is required'),
-  mode: z.enum(['analysis', 'write', 'auto']).default('analysis'),
+  mode: z.enum(['analysis', 'write', 'auto', 'review']).default('analysis'),
   format: z.enum(['plain', 'yaml', 'json']).default('plain'), // Multi-turn prompt concatenation format
   model: z.string().optional(),
   cd: z.string().optional(),
@@ -1176,7 +1176,8 @@ export const schema: ToolSchema = {
 Modes:
 - analysis: Read-only operations (default)
 - write: File modifications allowed
-- auto: Full autonomous operations (codex only)`,
+- auto: Full autonomous operations (codex only)
+- review: Code review mode (codex uses 'codex review' subcommand, others accept but no operation change)`,
   inputSchema: {
     type: 'object',
     properties: {
@@ -1191,8 +1192,8 @@ Modes:
       },
       mode: {
         type: 'string',
-        enum: ['analysis', 'write', 'auto'],
-        description: 'Execution mode (default: analysis)',
+        enum: ['analysis', 'write', 'auto', 'review'],
+        description: 'Execution mode (default: analysis). review mode uses codex review subcommand for codex tool.',
         default: 'analysis'
       },
       model: {

ccw/src/tools/cli-executor-utils.ts

@@ -223,7 +223,21 @@ export function buildCommand(params: {
 
     case 'codex':
       useStdin = true;
-      if (nativeResume?.enabled) {
+      if (mode === 'review') {
+        // codex review mode: non-interactive code review
+        // Format: codex review [OPTIONS] [PROMPT]
+        args.push('review');
+        // Default to --uncommitted if no specific review target in prompt
+        args.push('--uncommitted');
+        if (model) {
+          args.push('-m', model);
+        }
+        // codex review uses positional prompt argument, not stdin
+        useStdin = false;
+        if (prompt) {
+          args.push(prompt);
+        }
+      } else if (nativeResume?.enabled) {
         args.push('resume');
         if (nativeResume.isLatest) {
           args.push('--last');