AgentSkillsCN

recall

回忆与当前对话或特定查询相关的记忆

SKILL.md
--- frontmatter
name: recall
description: Recall memories relevant to the current conversation or a specific query
user_invocable: true
arguments:
  - name: query
    description: Optional search query. If omitted, infer context from the conversation.
    required: false

/openmem:recall

Recall relevant memories from the persistent memory store.

Instructions

  1. •

    Determine the recall query:

    • •If the user provided a $ARGUMENTS query, use it directly.
    • •If no query was provided, analyze the current conversation context and synthesize a focused search query from the key topics, entities, and decisions being discussed.
  2. •

    Call the memory_recall MCP tool with the query. Use top_k: 5 and token_budget: 2000 as defaults unless the user specifies otherwise.

  3. •

    Present the results to the user in a clear format:

    • •For each memory, show:
      • •Type and status
      • •Content (the memory text)
      • •Score (relevance score from 0-1)
      • •Confidence and strength
    • •If no memories are found, let the user know and suggest storing relevant information with /openmem:store.
  4. •

    If any recalled memories seem outdated or contradictory, mention this to the user and suggest using the memory management tools (supersede, contradict) to clean things up.