name: memory-orchestrator description: Orchestrate ClawXMemory retrieval using memory_search first and memory_get for exact record reads. Use when the user asks for historical context, project progress, timeline, or profile facts. homepage: https://github.com/OpenBMB/ClawXMemory/tree/main/clawxmemory metadata: {"openclaw":{"skillKey":"openbmb-clawxmemory","requires":{"config":["plugins.entries.openbmb-clawxmemory.enabled"]}}}

Memory Orchestrator

Use this skill when the task depends on conversation history, project status, timeline, or user profile facts.

Primary Path (Tool-first)

1. If the user is asking what memory exists or what ClawXMemory currently remembers, use memory_overview or memory_list instead of retrieval.
2. Otherwise call memory_search with the user's question.
3. Read intent, enoughAt, returned context, and refs.
4. If the returned context is sufficient, answer from it directly.
5. If you need exact verification, call memory_get with the smallest relevant id set from refs.
6. If memory_search returns weak or empty evidence, say so clearly and ask a targeted follow-up question.

Fallback Strategy

• Time-oriented question: prefer refs.l2 entries whose level is l2_time, then verify with memory_get({ level: "l2_time", ids }).
• Project-oriented question: prefer refs.l2 entries whose level is l2_project, then verify with memory_get({ level: "l2_project", ids }).
• Fact/profile question: start with memory_search, then inspect only the most relevant l1 or l0 ids if the answer needs precise support.

Tool Usage Notes

• Prefer concise query strings; avoid copying the entire user prompt unless necessary.
• Keep limit small, usually 4 to 8, unless the user asks for exhaustive history.
• For browse-only questions, use memory_overview / memory_list rather than memory_search.
• Use memory_get only after memory_search has given you the ids worth reading.
• If evidence conflicts, prefer the newest verified l1 / l0 record and say that the history appears inconsistent.

Minimal Example

User asks: "Where is this project now?"
1) memory_search({ query: "recent project progress", limit: 6 })
2) Read refs.l2 and choose the best l2_project id
3) memory_get({ level: "l2_project", ids: ["<l2-id>"] })
4) If exact wording still matters, read the linked l1 or l0 ids with memory_get

Example User Requests

• "What do you remember?" Use memory_overview or memory_list, not memory_search.
• "What project memory entries exist?" Use memory_overview or memory_list, not memory_search.
• "What is the latest progress on this project?" Start with memory_search, then verify the most relevant l2_project entry with memory_get if needed.

Guardrails

• Do not fabricate details not supported by retrieved memory.
• Prefer newer verified entries over older ones when there is conflict.
• Do not call memory_get with guessed ids.
• If retrieval is empty, state uncertainty clearly and ask a targeted follow-up question.