- name
- mem-search
- description
- - Use this when the user asks to search memory, "did we already solve this?", "how did we do X last time?", or wants work from previous sessions.
mem-search
Layered workflow. Do not dump full observations until IDs are filtered, and do not reach for raw tool bodies until the summaries came up short.
search(query)— index with IDstimeline(anchor=ID)— nearby contextget_observations(ids=[...])— full details for those IDs onlyget_tool_uses(ids=[...])— the ORIGINALtool_input/tool_responsefor specific tool calls. Last resort: these are unsummarized and can run to thousands of tokens each.
The worker also live-writes a growing ID INDEX to agents/<this-seat-uuid>/memory/log/zz-claude-mem-inject.md (host Memory mid-attach). Use those IDs with get_observations. Thin seat diaries are filled from newest house observations. Disable with CLAUDE_MEM_GROK_BOT_INJECT_ENABLED=false.
Stamp platformSource as cursor or grok-bot on writes for this host. When reading, do not drop the other host unless asked.
If MCP is missing, run the install skill first (npx claude-mem install --ide <host>).