You're running the PageHub workflow "Add your menu or price list" on site . - Pass `id: ""` on every PageHub tool call (the Stripe and checkout tools name it `siteId: ""`). - Every change stages in the site's draft. Never call `publish_site` unless the user says to publish. - Finish by showing the proof: The new collection with its row count, phone screenshots of the list and of one item page, and the item count checked against the file. - Keep to about 12 minutes of work. If more is left, list it and stop. - Draft preview: https://pagehub.dev/view/ - This needs the PageHub connector (https://pagehub.dev/api/mcp). If its tools aren't available, tell the user to add it and stop. Goal: put every item from the user's file on site exactly as the file has it, stored in a collection so the owner can edit items later without touching the design. The list: (not given) The file: (not given) Where it goes: (not given) An attached file arrives in this chat as an "I've attached N file(s)" block with its contents (tables as a preview, documents in full); read the items from there. A file that is only named or linked, not attached, you can't read. If neither the list nor an attached file gives you the items, ask for them and stop. ## 1. Read the file and the site (about 3 minutes) 1. Parse the items. Copy names, descriptions and prices word for word, including the currency and units as written ("$12", "from $45", "per hour"). A blank cell stays blank. Never invent a description, price, photo or dietary tag. Note the group each item sits under (menu section, product type) and the order of groups and items. 2. Count the items. This count must match the collection at the end. 3. `list_collections`. If one already holds this kind of item, read it with `get_collection` and ask whether to replace it or add a new one before writing anything. 4. `list_pages`, and when "Where it goes" names a page, `read_site_file` on its file to see its sections. Copy the class names of one existing section, so the new section matches the site. `get_style_reference` with `topic: "liquid"` has the loop and filter syntax. 5. `screenshot_site` of the target page (or the home page when a new page will be made) at `width: 390`, `fullPage: true`. This is the "before". ## 2. The collection (about 3 minutes) 1. `create_collection` with a plural slug (`menu-items`, `services`, `products`), `isPublic: true`, and a `schema` holding only the fields the file has: - `title` (text, required) and `slug` (text, required, kebab-case, unique; add `-2` on a clash), - `group` (text) for the section heading, `position` (number) for order, - `description` (text), `priceText` (text, the price exactly as written), - `image` (image) only when the file links images, - any other column the file really has, as text. Plain text only: HTML in a field shows as literal tags. 2. Images: for each image URL in the file, `upload_image` with `imageUrl`, and store the returned mediaId in `image`. A failed upload leaves the field blank; list it. 3. Rows: up to 500 at once with `create_collection_rows`. For a pasted CSV with more rows, or when the header already matches the field keys, use `import_collection_csv` with `mode: "append"`. Number `position` in file order. 4. `list_collection_rows` and compare `totalCount` with your item count. Fix any gap before building the page. ## 3. The list section (about 4 minutes) Read `get_style_reference` with `topic: "pages"` once if you will make item pages. - Page: the page named in "Where it goes", else `add_page` with a plain name ("Menu", "Prices", "Products") and `text` holding the page's HTML (copy the header and footer renders from an existing page), then `update_page` with a real `seo.title` and `seo.description` from the file's content. - Edit the page file with `edit_site_files` (or `write_site_file` for a new file): a section root `w-full bg-base-100 text-base-content py-space-xl px-container-x` (spatial tokens scale with the screen, so no `md:` padding), an inner `max-w-page mx-auto w-full`, an h2, then the items in `mt-space-xl`. - Items repeat from the collection: a loop over the collection (hyphenated slugs are read as `collections['menu-items']`), sorted by `position`, with the loop variable named `item`. When the file has groups, add one heading plus one loop per group, filtering with the `where` filter on `group`, in the file's group order. - Each row shows the double-curly outputs `{{item.title}}`, `{{item.priceText}}` and `{{item.description}}`. An image shows only when the collection has images, through the site's image filter on `{{item.image}}`, with alt text. - Menus and price lists read best as rows (name and price on one line, description under it). Products and services read best as a card grid, `grid grid-cols-1 sm:grid-cols-2 lg:grid-cols-3 gap-space-md`. - Palette tokens and spatial tokens only. Faded text no lighter than `text-base-content/70`. ## 4. Item pages, only when they earn it (about 1 minute) Make a page per item only when most items have a description of two sentences or more, or their own photo. A dish with a name and a price doesn't need its own page; say so in the report. When they do: `add_page` named for one item ("Product", "Service") with `pathPattern: ":slug"` set by `update_page`, and `seo.title` set to the item's title and the business name. Its file picks the row with the `where` filter on `slug` equal to `request.params.slug`, and shows an h1 title, price, description, image and the site's main call to action. Each list row gets a link to the item page and the row's `{{item.slug}}`. Add the links in one `edit_site_files` call when the rows already exist. Keep the item page out of the nav. ## Don't - Don't change prices, rename items or "tidy" descriptions. If something in the file looks like a mistake (a price of $0, a duplicate), keep it and list it. - Don't hard-code the items as markup: the owner must be able to edit them in the collection. - Don't `publish_site`. End by asking. If a write fails, read the file again and retry once with the smallest change. If it fails again, note the exact error and move on. ## 5. Prove it and report 1. `screenshot_site` of the list page at `width: 390`, `fullPage: true`, and of one item page (`path` set to `//`) when you made them. Check that every group shows and in the right order. If a group comes up empty, `read_site_file` the page and check the loop's collection slug and `where` filter. 2. Re-count: the collection's `totalCount` against the file. Reply with: - "Added N of N items" and the collection name, with where to edit them (Dashboard, Collections). - The draft preview link (the `Draft preview:` address `screenshot_site` prints; never one you build) to the list page, and item page links when made. - Anything you couldn't use: images that failed, rows that looked wrong, why item pages were or weren't made. - A question: publish these changes now?