recall

name	recall
description	Search and retrieve stored memories. Use when the user asks to recall, remember, look up past context, or when you need to find relevant information from previous conversations.
argument-hint	[search query or topic]
allowed-tools	mcp__agent-memory__memory_search, mcp__agent-memory__memory_context, mcp__agent-memory__memory_recall, mcp__agent-memory__memory_working_get

The user wants to retrieve memories. The query is:

$ARGUMENTS

Your job:

1. Determine the best retrieval strategy:
   • Semantic search (memory_search): use when the query is natural language or conceptual — "what does the user prefer for testing?", "anything about auth?"
   • Filtered recall (memory_recall): use when the query targets a specific project, agent, tag, or memory type — "all feedback memories", "memories tagged auth"
   • Task context (memory_context): use when loading context for a specific task — "I'm about to refactor the auth middleware"
   • WorkingContext (memory_working_get): use when the user asks what is currently in progress or how to resume the active project handoff.
2. Execute the appropriate MCP tool call(s)
3. Present results clearly:
   • Show the most relevant memories first
   • Include tags, project, and memory type for context
   • If a memory seems stale or potentially outdated, note that
   • Quote the memory content directly — don't paraphrase

If no results are found, say so and suggest alternative search terms.

If the query could benefit from multiple retrieval strategies, run them in parallel and merge the results.