Ask

A process-first wrapper around local LLM CLIs that keeps you from assembling raw commands yourself. You pick a provider (Claude, Codex, or Gemini), pass your prompt, and it handles the flags, versioning, and artifact persistence automatically. Everything gets written to `.omc/artifacts/ask/` with timestamps. The main value is enforcement: it wants you to route through `omc ask` instead of calling provider CLIs directly, which means less flag-juggling when you're bouncing between models or trying to capture outputs consistently. Good if you work with multiple LLMs locally and want a single interface that doesn't make you remember which flags changed in the last Codex update.