notion-onmyplate

name	notion-onmyplate
description	Check what's on the user's Notion plate by querying task databases, mentions, and waiting-on items. Use when the user asks about Notion tasks, follow-ups, what's assigned to them in Notion, what they're waiting on from others, or what they owe replies to in Notion.

Notion: What's On My Plate

This skill reports what needs the user's attention in Notion. Unlike gh-onmyplate / gl-onmyplate, there is no CLI — all queries go through the Notion MCP (mcp__claude_ai_Notion__notion-* tools).

Three buckets, each a separate query:

1. Assigned to me — Tasks where the user is the assignee in a known task database
2. I'm mentioned / need to follow up — Pages where the user is @mentioned recently (and may owe a reply)
3. Waiting on others — Tasks the user owns where someone else is the assignee or is listed in a Waiting on / Follow-up property

Configuration

Everything this skill needs is in 05 Meta/config.yaml under notion::

notion:
  self_name: "Jane Smith"          # used to match assignee / mentions
  task_databases:                  # databases that hold tasks
    - "abc123..."
    - "def456..."
  mention_lookback_days: 7         # how far back to look for mentions
  follow_up_wait_days: 3           # flag items waiting on others for N+ days

Read this file first. If task_databases is empty, skip bucket 1 and tell the user to add database IDs to the config.

Workflow

Step 1: Resolve the user's Notion identity

The notion_assignee and mentions fields in Notion are user IDs (UUIDs), not names. Resolve notion.self_name (from config) to the matching Notion user ID using:

mcp__claude_ai_Notion__notion-get-users

Cache the result for the session — don't re-query for each bucket.

Step 2: Query each bucket

Bucket 1 — Assigned to me

For each database ID in notion.task_databases:

mcp__claude_ai_Notion__notion-fetch with url=<database_url>

(or use notion-search scoped to the database). Filter the results to items where:

• The assignee / owner property contains the user's Notion user ID, AND
• The status is not Done / Archived / similar terminal state

Collect: id, url, title, status, due_date, last_edited_time, parent database name.

Bucket 2 — I'm mentioned / need to follow up

Search for pages that mention the user:

mcp__claude_ai_Notion__notion-search with query="@<self_name>"

(Or, if the MCP exposes a dedicated mentions lookup, prefer that.) Filter to items with last_edited_time within mention_lookback_days. Skip anything already covered by Bucket 1 (dedup by page ID).

For each hit, fetch just enough context to determine "does the user owe a reply?" — typically the last few lines after the user's mention.

Bucket 3 — Waiting on others

For each database in notion.task_databases:

• Find items where the user is the creator / owner (not assignee), AND
• An assignee or Waiting on / Follow-up property points to someone else, AND
• last_edited_time is older than follow_up_wait_days days ago (stale — time to nudge)

Collect: id, url, title, waiting_on_name, last_edited_time, due_date.

Step 3: Cross-reference the vault

For items across all three buckets, check if there's already a type: notion note in the vault:

obsidian vault={{VAULT_NAME}} base:query path="02 Areas/Notion.base" format=json

For each query result, filter by notion_id matching the MCP result. The output tells you:

• Has vault note — cite the vault alias in the briefing ([[notion-<slug>]])
• No vault note — suggest /notion-import <url> as a follow-up action

Step 4: Synthesize the briefing

Present a summary grouped by bucket. Use clear visual separation so the user can scan quickly:

Assigned to me:

• For each: title, status, due date (or "no due date"), parent database, URL
• Sort: overdue first, then by due date ascending, then last_edited_time descending
• Flag items overdue or due today

I'm mentioned (last N days):

• For each: title, who mentioned the user, 1-sentence summary of the surrounding context, URL
• Sort: most recent first
• Mark "likely needs reply" vs "FYI"

Waiting on others (idle ≥N days):

• For each: title, waiting on whom, days since last edit, URL
• Sort: oldest idle first

At the end, list any items that don't yet have a vault note and suggest: Consider /notion-import for: <URLs>.

Always present URLs as full clickable links (https://www.notion.so/...), not markdown-wrapped text.

When to Use This Skill

Trigger on prompts like:

• "What's on my Notion plate?"
• "What Notion tasks are assigned to me?"
• "What am I waiting on in Notion?"
• "Any Notion follow-ups I owe?"
• "Check Notion for pending tasks"

Do NOT use this skill to:

• Import a specific page — that's /notion-import
• Search Notion for arbitrary content — use mcp__claude_ai_Notion__notion-search directly
• Create or edit Notion pages — this skill is read-only

Known Limitations

• User ID resolution depends on notion-get-users returning the user with a matching display name. If the user's Notion display name differs from notion.self_name in config, adjust the config.
• Mentions search is approximate — notion-search with an @name query may pick up the name as plain text, not just mentions. Filter results heuristically (e.g. "does the page body actually contain an @-mention block near last_edited_time?").
• Property schema varies per database — "status" may be called Status, State, or a custom name. When fetching a new database for the first time, inspect the property schema and record the mapping to notion_status / notion_due / notion_assignee in session notes for reuse.
• Archived pages are typically excluded by the MCP by default, but confirm via archived: false in the response if rigor matters.
• No sync direction back to Notion — this skill is strictly read-only. Status updates in the vault (e.g. marking a type: notion note as done) do not propagate back to Notion.