Personal memory

Personal Memory

IMPORTANT: Start your response with a context preamble.

Call help_lookup(topic="personal-memory", mode="preamble") and display the returned preamble text as a blockquote. Then tell the user they can say "tell me more" for a step-by-step guide, or answer the scoping question below to proceed.

If the MCP call fails, fall back to:

Personal Memory — Deliberately capture and recall curated cross-session knowledge — decisions, preferences, and findings — organized by topic so it persists across projects and sessions.

How this differs from the other memory skills

attune has three memory surfaces; pick the right one:

You want to…	Use
Deliberately save a decision/preference/finding to recall later, by topic	this skill (personal_memory_*)
Store a classified or cross-agent pattern in project memory	memory-and-context (memory_*)
Pull findings auto-stashed from past sessions + the lessons corpus	recall (session_stash)

This skill is the curated, intentional store: content is polished and filed under a topic (and optional kind), not auto-captured.

Scoping

Before running, ask which operation the user needs:

1. Operation: "Capture something, recall by query, list your topics, or forget a topic?"
2. For capture: "What topic should it live under, and is this global or project-local?"
3. For recall: "What should I search for?"

Execution

Call the matching MCP tool:

personal_memory_capture

Save a decision, pattern, finding, or reference to personal memory.

Parameters:

• topic (required): The topic to file this under.
• content (required): The text to store (it is polished before storage).
• kind (optional): A sub-category within the topic (e.g. decision, preference, reference).
• project_local (optional, boolean): Store only for the current project instead of globally. Default is global.

personal_memory_capture(
    topic="release-process",
    content="Always verify the merge SHA contains the version bump before tagging.",
    kind="decision",
)

personal_memory_recall

Search personal memory with a natural-language query.

Parameters:

• query (required): What to search for.
• k (optional, integer): Max results to return.
• kind_filter (optional): Restrict to a single kind.

personal_memory_recall(query="how do I tag a release", k=5)

personal_memory_topics

List every topic currently stored. No parameters.

personal_memory_topics()

personal_memory_forget

Delete a topic, or one kind within a topic.

Parameters:

• topic (required): The topic to remove.
• kind (optional): Remove only this kind; omit to remove the whole topic.

personal_memory_forget(topic="release-process", kind="decision")

Output

• capture: confirm the topic (and kind) it was filed under.
• recall: present hits newest/most-relevant first, annotated with their topic and kind.
• topics: render the topic list, grouped if kinds are present.
• forget: confirm exactly what was removed.

Anti-Patterns

• DO NOT use this for auto-captured session findings — that is the recall skill's job.
• DO NOT use this for classified or cross-agent patterns — use memory-and-context (memory_store with a classification).
• DO NOT capture secrets or PII — personal memory is not an encrypted store; use memory_store(classification="SENSITIVE") for that.