memory-keeper

To save session context and compact history, trigger the extraction process to scan for preferences, decisions, and tasks, write them to memory.md, and automatically run /compact to reduce context window bloat.

Tracking decisions and deadlines requires extracting key facts into a persistent markdown file. This approach merges new findings into memory.md, ensuring important details survive long conversations while keeping the active context concise.

Yes, you can preserve user preferences and task status by extracting them into memory.md. The system reads and merges existing memory data to ensure durable context is maintained without overwriting previous session information.

Memory compaction does not delete previous notes because the process uses non-destructive merging. It reads the existing memory.md file and appends newly discovered items, preventing blind overwrites while reducing active context load.

Markdown persistence for managing context is needed during ongoing assistant sessions that accumulate durable preferences, decisions, and tasks. It prevents important facts from being lost in long conversations by capturing them into a persistent file.

Context window bloat happens when long conversations accumulate excessive history. You reduce it by running a context compaction command after saving key information to memory.md, which keeps the session concise without losing extracted essentials.