What problem does it solve? Coding agents forget everything between sessions and lose context within one. This Skill teaches an agent how to use the Atmina MCP server as durable shared memory, so decisions, findings, and procedures survive to the next prompt instead of being re-derived from training data. ## Core Features & Use Cases - Four memory reflexes: Recall (search before answering from training), Capture (write decisions and session notes with integrity-checked writes), Update (etag-guarded edits with conflict retry), and Consolidate (run → review → promote routine loop for a KB-wide current-state view). - Typed filtering and exhaustive listing: Use list_files for exhaustive audits and search for relevance-ranked discovery, with built-in and registered frontmatter pointer filters. - Index and navigation generation: Preview-then-apply workflows for generate_hub_index and generate_nav to maintain folder hub indexes and wiki/nav.yaml menus. - Use Case: Before answering "what did we decide about the auth flow?", the agent searches the team's Knowledge Base, reads the cited source file to verify evidence state, and answers from attested memory rather than guessing. ## Quick Start Ask the agent to orient with the Atmina MCP server by running health, whoami, and list_kbs, then search your Knowledge Base before answering any question about past team decisions.
