instagram-dashboard-prompt-windsor.mdReady to copy
# Ready-made prompt: live Instagram analytics dashboard (data-connected artifact) via Windsor.ai
This file holds a complete prompt you copy and paste into a Claude conversation (on the web or the Claude desktop app) to have Claude build a **professional, live Instagram analytics dashboard** for your account, with the same layout and features as the original dashboard, fully **in English**.
---
## Before you start (requirements)
1. **A paid Claude plan (Pro, Max, Ultra, Team or Enterprise)**, because connecting an artifact to connectors requires a paid plan. It works on the web and in the desktop app. On Team and Enterprise, the organization owner must turn on "Enable artifact connectors" in the organization settings.
- Artifacts created in Cowork before 19 August 2026 were "Live Artifacts" and cannot be edited in place. This prompt builds a new artifact.
2. **Windsor.ai added** as a connector in Claude (from the Connectors settings), with **your Instagram account linked** inside Windsor.
- Your account must be a **Business or Creator account** for all the numbers to appear (reach, views, audience).
- If you have fewer than 100 followers, some "daily new followers" numbers may not appear (an Instagram limit). Instagram also provides daily new follower data for the last 30 days only.
3. **Optional, for the radar (Watchout):** turn on Scheduled tasks and web search so the news and competitor section updates every morning.
## How to use it
Copy **all the text below the divider line** (from the word "Task" to the end of the file) and paste it into Claude as one message. Claude will ask you a few simple questions (such as competitor accounts), then build the dashboard.
> Note: the prompt is designed to find the name of the Windsor tool in your session automatically, because the tool name differs from account to account. You do not need to change anything technical.
---
# Task (copy from here down)
You are a front-end and data engineer. Build me **one self-contained artifact**: an **Instagram account monitoring dashboard** that pulls its data live from the **Windsor.ai** connector and shows it **fully in English**, left to right, in **light mode only**. Do not use the em dash or the en dash anywhere (not in the code and not in the text). Use Western digits (0 1 2 3 4 5 6 7 8 9) everywhere.
## Step 0: Find the account and the tool
1. Use the Windsor tool `get_connectors` to confirm that the **`instagram`** connector is linked, and read the account's `id` and `name` from it. Confirm the account name with me before you continue.
2. **Important:** the Windsor tool names in my session have a server prefix before `get_data`, `get_fields` and `get_connectors`. Find the actual full name of the `get_data` tool in this session and use it: (a) in all your calls, and (b) as the only connector tool you declare the artifact may call once it is published. Do not write a server ID from another session.
3. Use `get_fields` for the `instagram` connector to confirm the field names before pulling data.
## Step 1: Check the data shape (probe)
Call `get_data` once for each group below to confirm the shape before building (the connector is always `instagram`):
- **Profile:** `["username","name","followers_count","follows_count","media_count","biography","website"]`
- **Time series (last 30 days):** `["date","reach","views","total_interactions","likes","comments","shares","saves","follower_count_1d"]` with `date_preset: "last_30d"`
- **Posts (last 30 days):** `["media_id","media_type","media_permalink","media_thumbnail_url","media_url","media_caption","timestamp","media_like_count","media_comments_count","media_reach","media_views","media_saved","media_shares"]` with `date_preset: "last_30d"`
- **Audience:** three separate calls: `["audience_country_name","audience_country_size"]`, `["audience_gender_name","audience_gender_size"]` and `["audience_age_name","audience_age_size"]`
- If `get_data` returns the status "pending" with no rows, repeat the same call with the same parameters until the rows arrive.
Inside the artifact, read the rows from the `result` array of each response, and parse the response as JSON first if it arrives as text.
## Step 2: Dashboard sections (match each section exactly)
The dashboard has these sections in this order, and all of them update when the date range changes or the page reloads:
**A) Top bar:** a simple logo/name + the account name and @username + a date range picker (7 days / 30 days / 90 days) + the last updated time.
**B) KPI cards, 6 cards:**
- Followers = `followers_count` (current total)
- New followers = sum of `follower_count_1d` over the range. When the 90-day range is selected, Instagram provides this number for the last 30 days only, so show a small note "Last 30 days only" on this card.
- Reach = sum of `reach`
- Views = sum of `views`
- Interactions = sum of `total_interactions`
- Engagement rate = (sum of `total_interactions` ÷ sum of `reach`) × 100 %
For every card that can show a trend, compute a simple trend: split the range's rows in half, compare the sum of the second half with the first, and show ▲/▼ with a percentage.
**C) Two charts (Chart.js):** a daily "Reach and views" line chart, and a "New followers per day" bar chart.
**D) "Engagement breakdown" chart:** daily stacked bars (likes + comments + shares + saves) next to a "Profile" card (followers / following / posts + bio + website).
**E) "AI insights" section:** two cards:
- **Performance analysis:** use the artifact's ask-Claude feature from inside the page, and send it an object that summarizes the range's numbers (reach, views, new followers, best content type, best posting times taken from the hour of `timestamp`, and the top 6 posts). Ask for a short analysis in English (under 170 words) in simple Markdown.
- **Recommendations:** a second request to Claude from inside the artifact that returns two parts: "Your next step" (advice based on the numbers) and "Post ideas" (4 Reel/carousel ideas, each with a hook).
- Each card has a "Regenerate" button.
**F) "Top content" section (no carousel/slider):** tab buttons: **Most viewed / Most commented / Most shared / Most liked**. Each tab shows the **top 5 posts** as fixed cards in a grid, not a slider. Each card: a **post thumbnail** from `media_thumbnail_url`, otherwise `media_url`, with a type badge and rank number, the main value for the metric, small numbers (views/likes/comments/shares), a caption excerpt, the date, and a "View on Instagram" link.
- Artifacts block external images, including Instagram thumbnails, so always add a **fallback** (`onerror`) that shows a colored tile with the post type, so the layout does not break.
**G) "Audience" section:** three cards: "By gender" (doughnut: female/male/unknown), "By age" (bars for each age group), "Top countries" (horizontal bars for the top 10 countries with each country's flag and name).
**H) "Radar (Watchout)" section:** two cards:
- **AI news radar:** a list of the latest 5 AI news items, each with a suggested content angle and a source link.
- **Creators to watch:** cards for the accounts I choose, each with a short note about its latest content.
- Put this section's data in a script block marked by `<!--WATCHOUT_START-->` and `<!--WATCHOUT_END-->` that holds `window.WATCHOUT = { lastUpdated, news:[...], competitors:[...] }` so it can be updated automatically later (see Step 4).
**I) "All content" table:** a sortable table of every post in the range (date, type, link, views, reach, likes, comments, shares, saves, engagement).
## Step 3: Design rules (professional light mode)
- `dir="ltr"` and `lang="en"`, a warm light background (for example `#F7F5F0`), white cards with a light border and shadow, rounded corners, comfortable spacing.
- A system font (or one from Google Fonts only, the only external source allowed for fonts), with a comfortable `line-height`.
- A calm, professional color palette (dark navy for the main elements, one accent color for the CTA, and two secondary colors for the charts). You can adapt it to the account owner's brand.
- Format numbers with `toLocaleString("en-US")`.
- Labels to use:
Followers · New followers · Reach · Views · Interactions · Engagement rate · Reach and views · New followers per day · Engagement breakdown · Likes · Comments · Shares · Saves · Profile · Following · Posts · AI insights · Performance analysis · Recommendations · Top content · Most viewed · Most commented · Most shared · Most liked · Audience · By gender · By age · Top countries · Female · Male · Unknown · Radar · AI news radar · Creators to watch · All content · Date · Type · Post · Last updated · View on Instagram.
## Technical constraints for the artifact
- One self-contained HTML file: all CSS and JS **inline**, and images as `data:` when needed.
- The only charting library allowed is Chart.js, with exactly this tag:
```html
<script src="https://cdn.jsdelivr.net/npm/chart.js@4.5.0/dist/chart.umd.js" integrity="sha384-iU8HYtnGQ8Cy4zl7gbNMOhsDTTKX02BTXptVP/vqAWIaTfM7isw76iyZCsjL2eVi" crossorigin="anonymous"></script>
```
Build the table, the date range picker and the tabs with plain HTML/JS (no external libraries) so nothing depends on blocked CSS files.
- Set the charts to `responsive:true, maintainAspectRatio:false` and create them after their container is visible.
- Each section loads on its own with a loading state, and shows a friendly error message if a call fails.
- When the range changes (7/30/90), pull the time series and posts again and regenerate the AI insights.
- Write the whole file, then publish it as an artifact, and declare the full name of the Windsor `get_data` tool alone as the connector tool the page may call. Each viewer approves the connection the first time they open the artifact.
## Step 4: Update the radar automatically (optional but recommended)
- First ask me: **which Instagram accounts do I want to watch as competitors?** (enter the usernames). Save them in `window.WATCHOUT.competitors`.
- Then create a **daily scheduled task (7:00 AM my time)** that: (1) searches the web for the latest AI news and writes 5 items, each with a content angle and a link, (2) finds a short note about the latest content of each competitor, (3) reads the artifact file and replaces **only what is between the `<!--WATCHOUT_START-->` and `<!--WATCHOUT_END-->` markers** with a new `window.WATCHOUT` object (in the same order: `lastUpdated` in ISO format, `news`, `competitors`) without changing anything else in the file, then republishes the same artifact at the same link.
- Keep the tone of the ideas suited to an audience interested in making AI simple, with no em dash or en dash.
## Final output
After building: give me the artifact directly, summarize in two sentences the most important things it shows, and offer to turn on the morning radar update.
## Honesty note for the user
- Some audience numbers (gender/countries) may add up to less than the follower count, because Instagram classifies only part of the followers. Show them as distribution percentages, not exact counts.
- If thumbnails appear empty inside their box, it is because the network blocks them; the fallback tile keeps the layout intact. The alternative: pull the image links in the scheduled task and store them as `data:` if needed.
# End of prompt
Sources: the original prompt file, the Instagram connector field list in Windsor.ai, and Claude Help Center articles on artifacts and connectors. Checked November 2026.