AgentSkillsCN

collaborating-with-codex

在 Claude Code 中以非交互式方式调用 Gemini CLI，支持会话保存、上下文注入以及智能体角色注入。触发词：/gemini、gemini 协作、调用 gemini。

SKILL.md
--- frontmatter
name: collaborating-with-codex
description: Delegates coding tasks to Codex CLI for prototyping, debugging, and code review. Use when needing algorithm implementation, bug analysis, or code quality feedback. Supports multi-turn sessions via SESSION_ID and agent role injection from ~/.claude/agents/.

Quick Start

bash
python3 scripts/codex_bridge.py --cd "/path/to/project" --PROMPT "Your task"

Output: JSON with success, SESSION_ID, agent_messages, and optional error.

Agent Role Injection (NEW)

Inject agent roles from ~/.claude/agents/ to customize Codex behavior:

bash
# 使用 planner agent 角色
python3 scripts/codex_bridge.py --cd "/project" --agent planner --PROMPT "Plan the auth feature"

# 使用 security-reviewer agent 角色
python3 scripts/codex_bridge.py --cd "/project" --agent security-reviewer --PROMPT "Review auth.py for vulnerabilities"

# 列出可用 agents
python3 scripts/codex_bridge.py --cd "/project" --list-agents --PROMPT ""

Agent 注入优先级

  1. •--instructions - 直接传入指令字符串（最高优先级）
  2. •--instructions-file - 指定指令文件路径
  3. •--agent-file - 指定自定义 agent 文件路径
  4. •--agent - 从 ~/.claude/agents/ 加载 agent

Parameters

code
usage: codex_bridge.py [-h] --PROMPT PROMPT --cd CD
                       [--sandbox {read-only,workspace-write,danger-full-access}]
                       [--SESSION_ID SESSION_ID] [--skip-git-repo-check]
                       [--return-all-messages] [--image IMAGE] [--model MODEL]
                       [--yolo] [--profile PROFILE]
                       [--agent AGENT] [--agent-file AGENT_FILE]
                       [--agent-dir AGENT_DIR] [--list-agents]
                       [--instructions INSTRUCTIONS] [--instructions-file FILE]

Codex Bridge with Agent Role Injection

Core Options:
  --PROMPT PROMPT       Instruction for the task to send to codex.
  --cd CD               Set the workspace root for codex.
  --sandbox             Sandbox policy. Defaults to `read-only`.
  --SESSION_ID          Resume a previous session.
  --model MODEL         Model to use (e.g., gpt-5.2-codex).

Agent Role Injection:
  --agent AGENT         Agent name from ~/.claude/agents/ (e.g., 'planner').
  --agent-file FILE     Custom agent file path.
  --agent-dir DIR       Agent directory. Defaults to ~/.claude/agents/
  --list-agents         List available agents and exit.
  --instructions STR    Direct system instructions string.
  --instructions-file   Path to custom instructions file.

Other Options:
  --skip-git-repo-check Allow running outside a Git repository.
  --return-all-messages Return all messages including reasoning.
  --image IMAGE         Attach image files to the prompt.
  --yolo                Run without approvals or sandboxing.
  --profile PROFILE     Configuration profile from config.toml.

Multi-turn Sessions

Always capture SESSION_ID from the first response for follow-up:

bash
# Initial task with agent role
python3 scripts/codex_bridge.py --cd "/project" --agent architect --PROMPT "Design the API"

# Continue with SESSION_ID (agent role persists in session)
python3 scripts/codex_bridge.py --cd "/project" --SESSION_ID "uuid-from-response" --PROMPT "Add error handling"

Common Patterns

Planning with planner agent:

bash
python3 scripts/codex_bridge.py --cd "/project" --agent planner --PROMPT "Plan refactoring of auth module"

Security review:

bash
python3 scripts/codex_bridge.py --cd "/project" --agent security-reviewer --PROMPT "Audit payment.py"

Code review:

bash
python3 scripts/codex_bridge.py --cd "/project" --agent code-reviewer --PROMPT "Review recent changes"

Custom instructions:

bash
python3 scripts/codex_bridge.py --cd "/project" --instructions "你是 Rust 专家，专注于性能优化" --PROMPT "Optimize this function"

Debug with full trace:

bash
python3 scripts/codex_bridge.py --cd "/project" --agent build-error-resolver --PROMPT "Fix build errors" --return-all-messages

Process Monitoring (NEW)

使用 codex_monitor.py 监控和管理 Codex 进程：

bash
# 列出运行中的 Codex 进程
python3 scripts/codex_monitor.py --ps

# 查看最新会话的对话内容
python3 scripts/codex_monitor.py --session latest --messages 20

# 查看指定会话（使用 SESSION_ID）
python3 scripts/codex_monitor.py --session 019c08b7-8d56-7543 --messages 50

# 实时监控最新会话（Ctrl+C 停止）
python3 scripts/codex_monitor.py --watch

# 终止指定进程
python3 scripts/codex_monitor.py --kill 12345

# 终止所有 codex exec 进程
python3 scripts/codex_monitor.py --kill all

# JSON 格式输出（便于程序处理）
python3 scripts/codex_monitor.py --ps --json
python3 scripts/codex_monitor.py --session latest --json

Monitor 参数

参数说明
--ps列出运行中的 Codex 进程
--kill <PID|all>终止进程
--session <ID|latest>查看会话内容
--watch实时监控最新会话
--messages N显示消息数量（默认 20）
--jsonJSON 格式输出