What problem does it solve? Repeatedly answering the same AskUserQuestion prompts across Lintel skills wastes operator time when preferences are stable. This Skill lets operators tune each question to auto-decide, always ask, or ask only for irreversible decisions. ## Core Features & Use Cases - Per-question preference tuning: Set never-ask, always-ask, or ask-only-for-one-way for any question identified by a <skill>-<slug> question_id, persisted to a JSONL audit file. - Inline tuning: Reply with tune: <preference> directly after answering a question to record the preference without switching context, with profile-poisoning defense that only accepts tune: from the user's own chat message. - Compliance guardrails: Refuses never-ask for one-way-door questions such as push-to-main, secret rotation, and production data access. - Use Case: An operator who always approves the recommended scope reduction in plan reviews runs /plan-tune --set plan-eng-review-step0-scope-reduction never-ask so future sessions auto-decide that prompt. ## Quick Start Ask the agent to list your current question tuning preferences with /plan-tune --list, then set a preference for any question that fires too often.