memory-tool-usage

name	memory-tool-usage
description	This skill should be used when needing to recall earlier conversation details, after compaction occurs, when resuming a session, when details feel fuzzy, or when the user references something from earlier in the conversation. Provides guidance on effective use of the `recall` memory tool.

Memory Tool Usage

The recall tool is a conversation with your memory. Ask natural questions about what happened earlier — what you worked on, what you decided, what went wrong — and follow up to dig deeper.

recall(query="What was I working on before the refactor?")
recall(query="What error did we see in the JWT validation?")

Follow-up Questions

Each response includes a Session: <id> footer. Pass it back on follow-up queries so the conversation continues where it left off.

# Initial question
recall(query="When did we discuss the authentication bug?")
# Returns answer + "Session: abc123"

# Follow-up — picks up where the last answer left off
recall(query="What was the root cause?", session_id="abc123")

Time Constraints

Use time_start/time_end to focus on a specific period. Useful when a broad query reveals the relevant time window.

recall(query="When did we discuss the authentication bug?")
# Answer mentions it was around 2024-01-15T14:00:00

recall(query="What was the root cause of the auth bug?",
       time_start="2024-01-15T14:00:00", time_end="2024-01-15T16:00:00")

When to Use Memory Proactively

• After compaction: "What was I working on? What were the specific details?"
• When resuming a session: "What decisions did I make recently? What's still open?"
• Before making changes to code discussed earlier: "What did we decide about the API design?"
• When user references something from earlier: Don't guess, look it up
• When details feel fuzzy: If uncertain about specifics, query rather than assume

Tips

1. Ask natural questions: "What error did we see in the auth flow?" works better than keyword lists like "authentication error JWT"
2. Be specific about what you want to know: "What was the root cause of the login bug?" is better than "login bug"
3. Multiple focused questions beat one broad question: Ask targeted questions rather than one vague catch-all

Anti-patterns to Avoid

• Constructing keyword soup instead of asking a real question
• Proceeding with vague recollection when specifics matter
• Assuming the compaction summary captured everything important
• Guessing at implementation details instead of looking them up