メニュー
How to use OpenAI Codex via MCP tools for code generation, plan review, and collaborative implementation loops. Use when calling Codex for implementation, review, or any cross-model collaboration.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Smart size-based routing: Claude implements small changes (<=2 files, <=30 lines), Codex implements large changes, code-reviewer agent reviews all. Best for: executing a plan or task where you want automatic routing by complexity. Triggers on: /execute-codex, execute plan, implement with routing, run the plan.
Full TDD with smart routing: Claude writes tests first, Codex audits tests, then routes implementation by size (Claude small / Codex large), code-reviewer reviews. Best for: TDD on larger tasks where Codex should handle heavy implementation. Triggers on: /tdd-execute-codex, TDD with routing, test-driven execute, TDD large feature.
Claude implements code changes, Codex reviews via MCP with structured APPROVED/WARNING/BLOCKED verdicts. Best for: straightforward implementation where you want an external Codex review. Triggers on: /claude-codex, implement and review, build with Codex review, code with external review.
Create a detailed implementation plan with Opus-level reasoning, then Codex audits it for correctness, completeness, and security in a loop until approved. Best for: planning before coding — never modifies production code. Triggers on: /plan-codex, plan a feature, create implementation plan, design a system, architect this.
Full TDD workflow: Claude writes tests first (RED), Codex audits tests, Claude implements (GREEN), refactors (IMPROVE), Codex reviews final implementation. Best for: test-driven development where you want Codex reviewing both tests and code. Triggers on: /tdd-claude-codex, TDD with review, test-driven implement, tests first then implement.
| name | codex-mcp |
| description | How to use OpenAI Codex via MCP tools for code generation, plan review, and collaborative implementation loops. Use when calling Codex for implementation, review, or any cross-model collaboration. |
| Tool | Purpose |
|---|---|
mcp__codex__codex | Start a new Codex session. Returns a threadId + response. |
mcp__codex__codex-reply | Continue an existing session using threadId. |
Codex sessions are stateful via threadId. Always:
threadId from the first mcp__codex__codex callmcp__codex__codex-reply with that threadId for all follow-ups| Mode | When to Use |
|---|---|
read-only | Plan review, code audit, Q&A — Codex reads but cannot write files |
workspace-write | Implementation — Codex writes files in the project directory |
| Policy | When to Use |
|---|---|
never | Review-only tasks where no shell commands are needed |
on-failure | Implementation — auto-approve commands, intervene on errors |
Inject a persona or constraints into Codex without polluting the main prompt.
Always include: "Be concise. Do not include your reasoning process in the response. Only output the result."
The MCP server config uses 2>/dev/null at the process level.
Individual tool calls do NOT need stderr suppression — MCP returns structured JSON-RPC responses only.
See reference.md for detailed examples and patterns.