Notion Task Board — Reference for Ray

Durable reference doc so future Claude sessions can interact with the task board quickly and reliably without re-deriving the schema each time. Update this doc when the schema changes.

Locations

• Data source (collection): collection://ae46c40f-421d-4c94-924b-0a2b13de0756
• Title: “Task Board”
• Canonical URL: https://www.notion.so/8e2c0db1b29b4067881cbe68c987fc9d
• Read skill: ~/.claude/skills/check-board/SKILL.md (invoked by /loop 1h /check-board autonomous loop)

Schema

Status semantics

• Inbox — triage hasn’t happened. Founder tasks often land here; Ray should not pick these up until moved to To Do.
• To Do — ready to work. The autonomous loop picks from here.
• In Progress — actively being worked. Set this BEFORE starting work so the board reflects reality.
• Blocked — waiting on external input. Add a Notes entry explaining what’s needed.
• Done — completed. Add a Notes entry summarizing what was done.
• Archived — deprioritized or stale. The check-board skill auto-archives Ray-owned To Do items that have been stale 14+ days.

Owner semantics

• Ray — I work this autonomously. The check-board loop picks these up.
• Founder — Founder works this himself. I should NOT start these; just flag if stale (7+ days untouched).
• Both — Can start the parts I can do, mark Blocked on the parts needing founder input.

Priority ordering (check-board sort)

Check-board skill sorts tasks by:

1. Priority: Urgent → High → Medium → Low
2. Owner = Ray or Both first (Founder is skipped)
3. Status = To Do before Blocked

Pick top and work up to 3 per check cycle.

Task-creation conventions

Use these conventions so tasks are clearly autonomous-ready:

For a Ray-autonomous task:

• Owner = Ray
• Status = To Do
• Clear imperative title
• Notes contain:
  • What to do (1-3 sentences)
  • Acceptance criteria (bulleted, concrete — what does “done” look like)
  • Dependencies if any (on other board tasks by title, or external blockers)
  • Estimated context cost (S / M / L — helps the loop decide when to pick it up)
  • Skill to invoke if applicable (e.g. “Use /process-newsletter backfill batch:stratechery:1-20”)

For a decomposable multi-item job (like the newsletter backfill): Use the discovery → batch two-tier pattern (see ~/.claude/projects/-Users-ray/memory/feedback_queue_work_to_the_board.md):

1. Create ONE discovery task per source that scans, counts, and creates N batch tasks on the board.
2. Each batch task is independent and fits in one session’s context budget.
3. Batch tasks reference their source and range explicitly in the title (e.g. “Backfill Stratechery [1-20]”).

Creating tasks programmatically (Notion MCP)

Use mcp__claude_ai_Notion__notion-create-pages with:

• parent.type = "data_source_id"
• parent.data_source_id = "ae46c40f-421d-4c94-924b-0a2b13de0756"
• properties:
  • Task = title string
  • Status = “To Do” | “Inbox” | …
  • Priority = “Urgent” | “High” | “Medium” | “Low”
  • Owner = “Ray” | “Founder” | “Both”
  • Project = one of the project options
• content = Notion-flavored Markdown (description, acceptance criteria, etc.)

Example:

{
  "parent": {"type": "data_source_id", "data_source_id": "ae46c40f-421d-4c94-924b-0a2b13de0756"},
  "pages": [{
    "properties": {
      "Task": "Discover backfill: email@stratechery.com",
      "Status": "To Do",
      "Priority": "High",
      "Owner": "Ray",
      "Project": "Newsletter"
    },
    "content": "## What\nQuery Gmail for `from:email@stratechery.com`, determine total message count, and create batch tasks on the board of ~20 articles each (oldest first).\n\n## Acceptance\n- N batch tasks created, each titled `Backfill Stratechery [start-end]`\n- Discovery summary note filed at `06-reference/2026-04-11-backfill-discovery-stratechery.md`\n\n## Skill\nFollow discovery workflow in `~/.claude/skills/process-newsletter/SKILL.md`.\n\n## Context cost\nS (~5 minutes)"
  }]
}

Updating tasks

Use mcp__claude_ai_Notion__notion-update-page with the page ID returned from create-pages or from a fetch. Update Status before starting work, again on completion, and append to Notes with progress.

Search & query

• Search within the board: mcp__claude_ai_Notion__notion-search with data_source_url = "collection://ae46c40f-421d-4c94-924b-0a2b13de0756" and a query string.
• Full board scan: mcp__claude_ai_Notion__notion-fetch with the collection URL returns schema + pages (but can be large).

Related

• ../../.claude/skills/check-board/SKILL.md — the autonomous loop that reads this board
• feedback memory: Queue work to the board — the operating principle behind this
• ../01-projects/process-newsletter/README — example of a project using discovery+batch decomposition