| name | nmem-memory-working |
| description | Read the user's daily Working Memory briefing at session start or when recent priorities matter. This gives Google Antigravity cross-tool continuity without bloating the main prompt. Triggered by /nmem-memory-working. |
Read Working Memory
Always prefer native Model Context Protocol (MCP) or Nowledge FS paths for retrieval to prevent terminal command permission prompts and keep execution silent. Only fall back to CLI commands when MCP tools are disabled.
Preferred Retrieval Hierarchy
- Context Bundle (MCP): Call MCP tool
read_context_bundle to fetch startup identity, rules, and working memory.
- Working Memory (MCP): If context bundle is not needed, call
read_working_memory.
- Nowledge FS (MCP): Call
mem_fs tool with command: "cat", path: "/working-memory/working-memory.md" (or context).
- CLI Fallback: Only if MCP is unavailable, execute:
nmem --json wm read
Or context bundle fallback:
nmem --json context --source-app google-antigravity
If the runtime already knows the current project or agent lane, add --space "<space name>" to the CLI commands or set space_id in the MCP tools. Multi-agent orchestrators can set NMEM_AGENT_ID="<agent-slug>" before launching. Add NMEM_SPACE only when that whole run should override the identity's default space. Use NMEM_HOST_AGENT_ID only for advanced host-id aliases.
When to Use
- At session start
- When resuming work after a break
- When the user asks what they are focused on now
- When the current task clearly depends on recent priorities or active initiatives
- When updating focus areas or briefings: Use
wm-update entrypoint script.
Updating Working Memory In-Place
To update daily focus areas or briefing in Working Memory safely:
python3 hooks/nmem_entrypoint.py wm-update /path/to/updated_working_memory.md
Or supply raw string content:
python3 hooks/nmem_entrypoint.py wm-update --content "# Focus Areas\n- Priority 1..."
Usage Pattern
- Read Context Bundle or Working Memory once near the start of a session.
- If Context Bundle was already loaded and includes Working Memory, do not read Working Memory again.
- Reuse that context mentally instead of re-reading on every turn.
- Refresh or update only if the user asks, active focus areas change, or a long-running session needs it.
