What problem does it solve? AI agents have limited memory capacity, and valuable information like credentials, server configs, and user preferences gets lost when memory fills up. This Skill sets up a persistent, categorized markdown knowledge base so nothing important is ever deleted. ## Core Features & Use Cases - Categorized Knowledge Base: Creates a directory tree of markdown files under /home/projects/hermes-knowledge/ with categories for credentials, environment configs, habit rules, and memory archives. - Memory Archiving Workflow: When agent memory approaches capacity (~80%), old entries are archived to dated markdown files before removal, never deleted outright. - Lookup-First Rules: Enforces two permanent memory rules — archive before deleting, and search the knowledge base before asking the user to repeat information. - Use Case: A user shares an API key and server topology in chat. The agent files the key under 账号密码/ and the topology under 环境配置/, then weeks later retrieves them from the knowledge base when memory no longer holds them. ## Quick Start Set up a categorized knowledge base for me with folders for credentials, configs, habits, and memory archives, and record the two iron rules in memory.
