interviewing-baselinelisted

# Interviewing Baseline Protocol This skill defines the shared protocol that ALL interviewing/elicitation skills follow, the same way `review-toolkit:reviewer-baseline` unifies the reviewer agents. It exists so loop rules and the Decision Ledger contract live in exactly one place. **Canonical source notice:** this file is the single source of truth for the Decision Ledger schema and provenance enum. Every other site that restates them (`plan-interview/tools/ledger-lint.sh`, this plugin's `hooks/exitplan-ledger-gate.sh` via that lint, `review-toolkit:plan-reviewer`) carries a mirror marker and must be updated in lockstep when this section changes. ## Interview Loop Rules Rules for every interviewing turn, regardless of which skill is running: 1. **Explore first.** If a question can be answered by reading the codebase, the repo's docs, or ADRs (Grep/Glob/Read, or dispatching `review-toolkit:codebase-explorer` where the calling skill supports it), answer it yourself instead of asking. Asking the user a codebase-answerable question is a protocol violation. 2. **At most 2 questions per turn.** The user disengages otherwise. Related sub-choices may share one `AskUserQuestion` call, but the material decisions per turn stay ≤ 2. 3. **Attach a recommendation only when it's grounded** in the user's input, the codebase, or the repo's docs — cite the grounding. If nothing grounds an answer, ask plain — do not guess. (Reporter-owned facts — environment, frequency, business intent — a