用 Luna 子代理并行拆解 Codex 任务

one more thing: luna-worker.toml defines how the worker behaves. if you want luna_worker to become a natural part of your model's day-to-day workflow, update your AGENTS.md so the model knows when to use it. add this to your AGENTS.md ↓ When a task can be split into several independent and substantial subtasks, spawn one luna_worker per subtask and run them in parallel. Keep anything that takes only a few minutes in the main thread. Each worker runs in its own thread. Do not assume it can see the main thread conversation. Make every task description self-contained: files in scope, boundaries, and expected output. Independent read-only tasks can run in parallel. Any worker that writes files needs its own isolated worktree. Otherwise, run those workers sequentially. After all workers finish, the main thread checks each result against the acceptance criteria in the dispatch prompt before combining them. Re-dispatch anything that fails. If you dispatched multiple workers but only one ever runs, first check whether agents.max_concurrent_threads_per_session is set to 1 in config.toml. the same approach works for any other custom subagents you've configured.

Codex tip: tell Sol to configure 𝗟𝘂𝗻𝗮 𝗠𝗮𝘅 𝗮𝘀 𝗮 𝘀𝘂𝗯𝗮𝗴𝗲𝗻𝘁. 𝗽𝗿𝗼𝗺𝗽𝘁: create a custom agent named luna_worker at ~/.codex/agents/luna-worker.toml. use these settings: model = "gpt-5.6-luna" model_reasoning_effort = "max" give it a description and instructions for bounded delegated work. preserve the rest of my config. validate it against my installed Codex version, show me the diff, then use luna_worker for subagent tasks.