MemoryBridge Memory Manager

Problem

Agents can lose decisions, unfinished tasks, and operational context across long conversations and sessions. memory-manager treats memory as files rather than chat history: daily logs, long-term memory, and bridge snapshots are kept separately so later sessions can recover context.

How It Works

The skill exposes commands through memory_cli.py: log appends entries, search performs keyword lookup, recall extracts context, summary builds recent summaries, merge folds daily records into MEMORY.md, clean archives old logs, bridge restores cross-session summaries, and stats reports memory metrics. It also defines agent behavior rules: trigger immediate logging for decisions, deployments, confirmations, or completed operations; proactively run bridge in new sessions to surface recent topics, open tasks, and key decisions; and use check to inspect log freshness, merge status, bridge snapshots, and daemon state. Entries use timestamps such as ## HH:MM topic, with markers for emphasis, completion, and risk.

Scope And Caveats

This workflow fits agents that need continuity across deployment, configuration changes, or project work. For short conversations, frequent logging may create redundant entries. When multiple agents share a --workdir, path isolation should be verified to prevent MEMORY.md or bridge snapshots from overwriting each other.