What problem does it solve? Long-running AI workflows accumulate user preferences and constraints that should persist across sessions, but without curation they become cluttered with inferred lessons, task outcomes, and duplicated settings. This Skill governs how standing memory entries are reviewed, approved, scoped, and maintained so only deliberate user preferences persist. ## Core Features & Use Cases - Approval-gated curation: Shows the exact content and scope of any memory addition or edit and requires explicit user approval before writing. - Scoped memory management: Supports common memory, per-Project memory via project_id, and read-only wildcard scope, with per-scope indices starting at 1. - Safe edit and delete operations: Uses exact project_id, id, and index_revision values from listings, with stale-version recovery by re-reading the list instead of blind retries. - Use Case: A user asks to remember that they prefer concise commit messages. The Skill drafts the exact statement, confirms the target Project scope, obtains approval, and writes it to standing memory. ## Quick Start Ask the assistant to remember a specific preference, such as requesting that it store your preferred code review style in the current Project's memory.
