What problem does it solve? Pending questions (asks) waiting for human judgment can pile up unnoticed in the calm ask store, and structurally identical questions may repeat for weeks without anyone recognizing the pattern. This Skill watches the ask store continuously and surfaces those recurring decision patterns so they can be resolved once as a general rule instead of answered one by one. ## Core Features & Use Cases - Event-driven monitoring: Polls the asks table every 10 seconds via a Monitor background script and only reacts when the open-ask count, last_seen_at, or id set actually changes. - Isomorphic pattern detection: On each change, re-reads all open asks and lets the LLM judge whether questions share the same decision structure even when their topics differ. - Meta-ask filing with deduplication: Checks existing meta-asks in the database before filing a new one (kind="meta") asking whether that question type can be self-decided going forward. - Use Case: You notice the same kind of design question keeps landing in your ask inbox. Invoke /ask-watch, and it watches the store, detects three structurally identical open asks, and files a single meta-ask proposing a standing rule. ## Quick Start Ask the AI to monitor the ask store for recurring question patterns by saying "ask storeを監視して" or invoking /ask-watch.