hyperbrief-trigger-check

# Hyperbrief Trigger Check — the escalation gate You are about to ask the user a decision question, OR Superscalar just opened a write/deploy/send lane, OR an inbound `DECISION_REQUEST` arrived. **Before composing the message**, run this 30-second rubric. ## 1. Trigger detection Run this skill if ANY of these apply: - Next assistant message will contain `괜찮을까요` / `할까요` / `should we` / `which (one|option)` / `approve` / `confirm` / `choose between` / `OK to` / similar decision-soliciting patterns. - Superscalar fan-out just opened a `write`/`deploy`/`send` lane (Superscalar's irreversibility barrier touches a default-forbidden action). - Inbound A2A is `CUSTOM/DECISION_REQUEST` requiring response. - About to call `AskUserQuestion` tool. **Skip** if: - All fan-out lanes are read-only (no side effects). - This is a continuation of an already-decided plan within the same approved Phase (Constellation §4: "no confirming planned dispatch"). ## 1.5 Audience-profile command routing (v0.6+) Independent of the escalation rubric, certain user inputs are **audience-profile commands** that must be forwarded to the `hyperbrief` skill (which owns `AudienceProfileFallback` IR population) for parsing + persistence, regardless of whether a decision is being made. Recognized command shapes (parse BEFORE the escalation rubric; if matched, route AND continue): - **Tone profile**: `L<a>.<b>.<c>` (e.g. `L1.1.1`, `L2.2.2`) — sets `audience_profile_fallback.tone_profile`. - **term_pairing c...