Miremo Skill

Prerequisite: The Miremo MCP server must be connected in OpenClaw before this skill's tools are available. See the README for setup instructions.

Miremo is an AI note-taking tool. It stores the user's memos, documents, supertags (topic tags with rich content), and AI-extracted knowledge graph entities. All tools only access the currently authenticated user's own data.

Research Strategy

When the user asks about their knowledge, choose one of three research modes and follow it through completely before giving a final answer. Never give up after one failed search.

Mode A — Browse (user wants an overview)

Triggers: "what notes do I have", "show me my recent notes", "what have I been writing about", "give me an overview"

Steps:

1. list_memos(page_size=30) to sample recent memos
2. list_supertags() to see all topic tags at a glance
3. Synthesize a structured overview from both results

Mode B — Topic Research (most common)

Triggers: "what do I know about X", "find notes about X", "anything related to X", "my thoughts on X"

Steps:

1. global_search(query="X") — cross-type overview (memos + docs + supertags)
2. search_memos(query="X", search_type="semantic") for deeper semantic matches
3. If a relevant supertag appears: list_supertags(q="X") to expand via that tag
4. For people, concepts, or events that matter: list_entities() → get_entity_graph(entity_id) to explore relationships

Iteration rules (critical):

• If initial search returns few results, retry with synonyms, English equivalents, or split keywords before concluding "nothing found"
• Lower similarity_threshold to 0.25–0.35 on the second attempt
• hit_text is only a snippet — do not draw conclusions from it alone; use global_search to confirm scope across types
• Declare "no relevant notes found" only after at least 2–3 distinct search strategies all return empty

Mode C — Exact Lookup

Triggers: "do I have a note about X", "find the exact note where I wrote Y", "the note titled Z"

Steps:

1. search_memos(query="Y", search_type="full_text") for precise phrase matching
2. If not found, fall back to Mode B with semantic search

Available Tools

Search Tools

search_memos — Search memos by keyword or natural language.

• query: search text, e.g. "Python async programming notes"
• limit: max results (default 10, recommend ≤ 20)
• search_type: "hybrid" (default, most comprehensive) | "semantic" (natural language) | "full_text" (exact match)
• Returns: id, hit_text, similarity_score, created_at, modified_at

global_search — Cross-type search across memos, documents, and supertags.

• query: search text
• limit: max results per type (default 10)
• include_memos / include_documents / include_supertags: toggle each type (all true by default)
• Returns items with type ("memo" / "document" / "supertag"), id, title, description, score

List Tools

list_memos — Paginated list of memos.

• page_index, page_size (default 20), q (optional fuzzy filter)
• Returns: { items: [{id, outline_preview, created_at, ...}], total, page_size, page_index }

list_supertags — List topic supertags.

• q (optional filter), page_index, page_size (default 50)
• Default page size of 50 usually retrieves all tags in one call

list_documents — List uploaded documents (PDFs, files).

• q (optional name filter), page_index, page_size (default 20)
• Returns: document_id, name, created_at, modified_at, status

Create Tools

create_memo — Create a new memo.

• content: memo body, multi-line supported. First line becomes the top-level title; subsequent lines become sub-content.
• Returns: { memo_id: "<new UUID>" }
• After creation, vectorization and knowledge graph updates run automatically in the background.

Knowledge Graph Tools

list_entities — List AI-extracted knowledge graph entities.

• entity_type (optional): person, concept, place, organization, event, etc.
• page_index, page_size (default 20)
• Omit entity_type to get all types mixed

get_entity_graph — Get an entity's 1-hop relationship graph.

• entity_id: obtain via list_entities first
• Returns: { entity: {entity_id, name, entity_type, summary}, related_entities: [...], relationships: [{source_entity_id, target_entity_id, description}] }

When to Use Which Tool

Recommended Workflows

Answer "What do I know about X":

1. global_search(query="X") for a cross-type overview
2. If more detail needed: search_memos(query="X", search_type="semantic")
3. If a relevant supertag exists: list_supertags(q="X") to expand further

Help user record new information:

1. Confirm intent, then call create_memo(content="...")
2. Tell the user the note was created and show the memo_id
3. Structure content well: first line = core topic, subsequent lines = details

Explore user's knowledge structure:

1. list_supertags() for a thematic overview
2. list_entities() to understand main knowledge graph nodes
3. get_entity_graph(entity_id) for entities of interest