optionslisted

> `<cai>` is the cai-codex command line that `$setup` wrote into your instructions (the cai-codex block in AGENTS.md). `<cai-root>` is what `<cai> --root` prints. Lay out options someone can actually choose between: $ARGUMENTS Two entrances, one procedure. **Before** — a reply is about to offer two or more ways forward. **After** — one already went out and the reader could not act on it; rewrite that message, never restate it. The always-on half of this lives in `rules/option-explainer.md` and applies whether or not this skill was invoked. This file is what does not fit in that file's line ceiling: the skeleton, the dimension library, and worked examples. If `$ARGUMENTS` is empty and the previous message holds no option list, say so and stop. Do not invent a set of options to expand. ## Which reference to read | Situation | Read | |---|---| | You need the output skeleton | `<cai-root>/skills/options/references/template.md` | | You are unsure what to compare on | `<cai-root>/skills/options/references/dimensions.md` | | You want the worked good and bad pair | `<cai-root>/skills/options/references/good-bad.md` | Read the one you need, not all three. ## Procedure 1. **Is there a choice at all?** One reasonable way forward — say so and stop. An option that is plainly worse — name it, say why it is excluded, leave it out. Never pad to three. 2. **Missing a fact the choice turns on?** Ask one question — one — and stop there. Do not list options against a guess. 3.