What problem does it solve? Vague statements like "I should know more about X" are hard to act on directly, and expanding them inline pollutes the main conversation context with exploratory reasoning. This Skill turns a fuzzy gap into a sharp, actionable pair of concrete questions plus validated source candidates, while keeping all expansion reasoning inside a sub-agent's window. ## Core Features & Use Cases - Delegated expansion: A single sub-agent parses the seed gap, generates 1-3 concrete questions, and proposes source candidates, so the caller's context stays clean. - Source validation: Candidate sources are spot-checked with Glob and Read (capped at 3 files, 200 lines each) so only plausible, existing sources are proposed. - Confirm/edit/refuse gate: The caller sees only the proposed question-source pair and decides whether to confirm, edit, or refuse before any downstream work. - Use Case: Before modifying a rate limiter, you say "prime me before I touch the rate-limiter". The Skill returns a sharp question like "How does the current rate-limiter handle burst traffic?" paired with the relevant source files, ready to feed into /prime-fetch or /prime-sweep. ## Quick Start Invoke /prime-expand with a vague gap statement such as "expand on the cache layer" and confirm or edit the proposed question and source pair it returns.
