Standard test scenarios

Run these after connecting. Expect the AI to call MCP tools (not guess content).

1. Catalog

List my Guides notebooks.

Pass: Returns titles and slugs (e.g. shopping preferences, meal plan, on-buddhism).

2. Read structure

Show me the structure of my on-buddhism guide - which pages are empty?

Pass: Uses guides_outline or guides_get; lists item_id values and placeholder pages.

3. Fill placeholder (no duplicate page)

Read on-buddhism, find the Theravada Buddhism page, and fill it with a concise overview.

Pass: Uses guides_fill_page with the existing item_id - does not append a second Theravada page.

4. Scaffold + fill

Scaffold a Weekly Meal Plan with sections Breakfast, Lunch, Dinner and a page called This Week under Dinner.

Pass: One guides_scaffold call; returns sections_created, pages_created, item ids.

5. Cross-guide planning

Read my shopping preferences and meal plan guides. Suggest dinners for next week and add a This Week page to the meal plan.

Pass: Reads two guides, writes one new page (append only if title does not exist).

6. Cleanup duplicate

If on-buddhism has two pages with the same title, delete the duplicate.

Pass: guides_outline → guides_delete_item on the extra id only.

7. Merge (optional)

Merge my two overlapping notes pages into one and delete the source.

Pass: guides_merge_items.

Report failures with: platform (ChatGPT / Cursor), prompt used, and whether guides_catalog worked.