# `euporia/` — `/elicit`

**Responsibility under the Northstar** (root `AGENTS.md`). The protocol here turns an intent the person can feel but not yet state into the decisions it actually turns on. The AI reads the person's own material — codebase, rules, past sessions — for those decision points and asks about them, cycle by cycle, until the intended endpoint is settled.

**Boundary.** `skills/elicit/SKILL.md` is the contract for `/elicit`; it names any supporting file in its `references/` to read where that file applies. `README.md` and `README_ko.md` explain the protocol to people, and `.claude-plugin/` and `.codex-plugin/` carry what a host needs to list and install it; what the protocol does once invoked is set in the `SKILL.md`.

**Next.** `skills/elicit/SKILL.md`.