Second Brain Setup

intent

this skill scaffolds a notion second brain workspace in one shot. it creates a Second Brain parent page with two child databases, Reading List and Inbox, using only notion mcp tools. run this once to bootstrap your second brain, or run it again to repair a broken setup (missing databases, wrong parent, etc.). the skill is strict: it creates only those three objects, no methodology pages, no emojis, no body content. if a database already exists under the right parent, the skill adopts it as-is without modifying schema.

inputs

• Notion integration connection: the skill requires an active notion integration (e.g., IronClaw) with read-write access to at least one notion page. set this up in your notion workspace: create a page, open its share menu (three dots), select "Connections", add the integration by name, and grant access. the integration must be shared with a top-level or accessible workspace page before you run this skill.
• notion-search mcp tool: used to scan for existing Second Brain, Reading List, and Inbox objects. requires query and query_type ("internal" only).
• notion-create-pages mcp tool: used to create the Second Brain parent page. requires parent (page_id) and properties (title only).
• notion-create-database mcp tool: used to create Reading List and Inbox databases. requires parent (page_id), title, and properties (schema).
• notion-move-pages mcp tool: used if Reading List or Inbox exist but have the wrong parent. requires page_ids array and parent (page_id).
• edge case: no pages accessible: if the integration has not been shared with any page, the skill stops and tells you to share a page first.
• edge case: rate limits: notion-search is capped at 30 calls per minute. this skill makes at most 3 searches (one for Second Brain, one for Reading List, one for Inbox), so rate limits are not a concern.
• edge case: auth expiry: if the integration token is revoked or expired, all tool calls fail immediately. the skill stops and reports the failure.
• edge case: network timeout: if any tool call times out (>30 sec), the skill retries once with the same payload. if the retry fails, the skill stops and tells you to check your connection.

procedure

step 0 - prerequisite: find an accessible parent page

the notion integration can only create child pages under a page it has already been shared with. there is no workspace root creation path.

call notion-search with payload:

{"query": "", "query_type": "internal"}

examine the results:

• if you see a page titled exactly Second Brain, save its page_id as SECOND_BRAIN_ID and jump to step 2 (database adoption).
• otherwise, pick any visible page from the results as INTEGRATION_PARENT_ID. prefer a top-level page (no parent, or with a name like Home, Workspace, IronClaw, etc.).
• if the search returns zero results, stop and output the exact message: No pages accessible to the integration. In Notion, create a page (e.g. "IronClaw") and share it with the integration (... → Connections → IronClaw), then run setup again.

input: notion search with empty query. output: list of pages. saves SECOND_BRAIN_ID or INTEGRATION_PARENT_ID.

step 1 - create the second brain parent page (conditional)

only run this step if SECOND_BRAIN_ID was not found in step 0.

call notion-create-pages with exactly this payload:

{
  "pages": [
    {
      "parent": {"type": "page_id", "page_id": "<INTEGRATION_PARENT_ID>"},
      "properties": {"title": "Second Brain"}
    }
  ]
}

do not include any other fields, even empty ones: no content, icon, cover, template_id, children, tags, or description. the pages[0] object must have exactly two keys.

input: INTEGRATION_PARENT_ID from step 0. output: SECOND_BRAIN_ID (the page_id of the created page), and SECOND_BRAIN_URL (the public url).

step 2 - find or adopt reading list database

call notion-search with:

{"query": "Reading List", "query_type": "internal"}

branch on the result:

• database found, parent is already SECOND_BRAIN_ID: do nothing. record READING_LIST_ID and READING_LIST_URL. move to step 3.
• database found, parent is different: call notion-move-pages with page_ids: [<READING_LIST_ID>] and parent: {type: "page_id", page_id: "<SECOND_BRAIN_ID>"}. record the returned READING_LIST_ID and READING_LIST_URL. move to step 3.
• database not found: call notion-create-database with:

{
  "parent": {"type": "page_id", "page_id": "<SECOND_BRAIN_ID>"},
  "title": [{"type": "text", "text": {"content": "Reading List"}}],
  "properties": {
    "Name": {"title": {}},
    "Type": {"select": {"options": [
      {"name": "Book"}, {"name": "Article"}, {"name": "Podcast"}, {"name": "Video"}
    ]}},
    "Status": {"select": {"options": [
      {"name": "Not started"}, {"name": "In progress"}, {"name": "Done"}
    ]}},
    "Score": {"number": {"format": "number"}},
    "Author": {"rich_text": {}},
    "Link": {"url": {}},
    "Review(Sum up)": {"rich_text": {}},
    "Tags": {"multi_select": {"options": []}}
  }
}

record READING_LIST_ID and READING_LIST_URL. move to step 3.

input: SECOND_BRAIN_ID from step 1 or step 0. output: READING_LIST_ID, READING_LIST_URL, and action taken (created/adopted/moved).

step 3 - find or adopt inbox database

call notion-search with:

{"query": "Inbox", "query_type": "internal"}

branch identically to step 2:

• database found, parent is already SECOND_BRAIN_ID: record INBOX_ID and INBOX_URL. move to step 4.
• database found, parent is different: move it under SECOND_BRAIN_ID. record INBOX_ID and INBOX_URL. move to step 4.
• database not found: call notion-create-database with:

{
  "parent": {"type": "page_id", "page_id": "<SECOND_BRAIN_ID>"},
  "title": [{"type": "text", "text": {"content": "Inbox"}}],
  "properties": {
    "Note": {"title": {}},
    "Type": {"select": {"options": [
      {"name": "Thought"}, {"name": "Task"}, {"name": "Resource"}, {"name": "Reference"}
    ]}},
    "Processed": {"checkbox": {}},
    "Date": {"date": {}}
  }
}

record INBOX_ID and INBOX_URL. move to step 4.

input: SECOND_BRAIN_ID from prior steps. output: INBOX_ID, INBOX_URL, and action taken (created/adopted/moved).

step 4 - report success

reply to the user with exactly this format (substituting urls and actions):

Second Brain ready ✓
• Second Brain page: <SECOND_BRAIN_URL>
• Reading List (created|adopted|moved): <READING_LIST_URL>
• Inbox (created|adopted|moved): <INBOX_URL>

If captures don't work, make sure the IronClaw integration is connected to the Second Brain page (... → Connections → IronClaw). Sharing the parent cascades to children.

input: all urls and actions from steps 1, 2, 3. output: formatted success message.

decision points

• if step 0 search returns zero results: stop immediately. the integration has no accessible pages. tell the user to create and share a page first, then retry.
• if Second Brain page already exists in step 0: skip step 1 entirely. use the existing SECOND_BRAIN_ID for steps 2 and 3.
• if Reading List exists but has wrong parent in step 2: move it. do not create a duplicate.
• if Reading List exists with its own schema different from the spec: adopt it as-is. do not modify its properties.
• if Inbox exists but has wrong parent in step 3: move it. do not create a duplicate.
• if any notion mcp tool returns error code -32602 (invalid params): strip all non-required fields from the payload, retry once, and report the error if retry fails.
• if any tool times out or returns network error: retry once with identical payload. if retry fails, stop and tell the user to check their notion connection.

output contract

success is three notion objects, created or adopted, and confirmed in a formatted message:

• Second Brain page: type page, title exactly "Second Brain", no body content, no icon, no cover, parent is INTEGRATION_PARENT_ID (from step 0), accessible at <SECOND_BRAIN_URL>.
• Reading List database: type database, title exactly "Reading List", parent is Second Brain, properties include Name (title), Type (select), Status (select), Score (number), Author (rich_text), Link (url), Review(Sum up) (rich_text), Tags (multi_select). accessible at <READING_LIST_URL>.
• Inbox database: type database, title exactly "Inbox", parent is Second Brain, properties include Note (title), Type (select), Processed (checkbox), Date (date). accessible at <INBOX_URL>.

all urls are public notion sharing links.

outcome signal

the user knows the skill worked when:

1. they see the success message: "Second Brain ready ✓" with three urls.
2. they open the notion workspace and see a Second Brain page.
3. clicking into Second Brain shows two child databases: Reading List and Inbox.
4. Reading List has the correct columns (Name, Type, Status, Score, Author, Link, Review, Tags).
5. Inbox has the correct columns (Note, Type, Processed, Date).
6. if the user shares the Second Brain page with the integration again (Connections menu), they can now use capture skills to add items to Reading List and Inbox without permission errors.

if the user runs the skill a second time and all three objects already exist with correct parents, the output is "Second Brain ready ✓" with the same urls and action labels read "adopted".