maestro-companion

> **Agent timeout**: `spawn_agent` 异步执行且无内置超时 — 除明确短任务外一律 `spawn_agent` 后立即 `wait_agent({ timeout_ms: 3600000 })`（上限 1 小时）阻塞等待，绝不依赖 30000 默认值；`timed_out: true` 且 Agent 未完成时再次 `wait_agent` 续等，不丢弃。批量场景使用 `spawn_agents_on_csv({ max_runtime_seconds: 3600, ... })`。 <required_reading> @~/.maestro/workflows/run-mode.md @~/.maestro/workflows/codex-run-mode.md </required_reading> <purpose> Minimal-run execution channel. Full LLM capability with minimal protocol: one `session start` + one `session done`, evidence appended to `{run_dir}/evidence/companion-log.md`. Use when: - Intent is mechanically clear (no design decisions needed; file count irrelevant) - No typed artifact consumed by downstream steps - No gate/verdict needed for lifecycle tracking </purpose> <context> $ARGUMENTS — intent text + optional flags. | Flag | Effect | |------|--------| | `-y` | Skip confirmation, execute directly | | `--note <text>` | Append note to active run's evidence log | | `--log <run_id>` | View evidence log for a specific run | | `--promote` | Promote run insights to spec/knowhow | Mode detection: `--note` → note | `--log` → log | `--promote` → promote | intent → execute | empty → ask </context> <invariants> 1. Only `session start` + `session done` — no prepare/brief/check/gates 2. Evidence is append-only, non-formal (never enters gates or artifact registry) 3. `--promote` delegates to `maestro-spec add` / knowhow capture, never writes directly 4. No auto-orchestration — executes directly, ne...