What problem does it solve? It consolidates the scattered logic for writing memories into ki-search into one single source of truth, eliminating duplicated format constraints, availability checks, scope resolution, and deduplication rules across skills like expert-team, strong-relation, debug, and code-review. ## Core Features & Use Cases - Seven Memory Type Strategies: Dispatches writes for topic memories, API interfaces, data flows, decisions, error library, strong relations, and pending changes, each with its own reference policy file. - Write-Time Hard Constraints: Enforces plain-text formatting (no YAML, headings, or tables), mandates ki_sync_relation APIs over ki_store, and controls vectorization via the vector parameter. - Deduplication and Degradation: Queries existing group entries before writing to update instead of duplicating, and gracefully skips writes when ki is unavailable without blocking callers. - Use Case: After expert-team builds a module expert, call this skill to write the topic memory, interface info, and data flow atoms into ki-search with correct groups, tags, and dedup checks. ## Quick Start Ask the AI to record a decision or error into ki memory, for example: record this decision about using polling instead of webhooks into ki memory.
