What problem does it solve? AI coding agents forget project-specific rules, architectural decisions, and test commands between sessions, forcing managers to repeat constraints and causing hallucinated violations of established conventions. ## Core Features & Use Cases - Persistent Memory Storage: Saves Manager-stated rules and constraints into logical namespaces (testing, database, deployment) using the store_memory MCP tool. - Proactive Context Retrieval: Searches and injects relevant memories at the start of every implementation task during the Context Phase. - Supersession & Safe Deletion: Detects outdated rules via search_memory, auto-supersedes same-topic entries, and requires explicit Manager approval before other deletions. - Auto-Generated Memory Index: Maintains a Markdown index at .opencode/memory/index.md for compact two-step retrieval without loading all memories. - Use Case: When a Manager says "never use Prisma push in this project", the skill stores it under the database namespace and retrieves it automatically before any future migration task. ## Quick Start Tell the agent to remember that this project always uses Jest with the --runInBand flag, and it will store the rule and recall it in future testing tasks.
