Catalog
stripe/stripe-bfcm-dashboard

stripe

stripe-bfcm-dashboard

Builds a Black Friday and Cyber Monday (BFCM) season dashboard of payments, customers, and recurring revenue from the connected Stripe MCP account, as a hosted dashboard or artifact where the agent's tools support one, otherwise as one HTML file. For the 2026 BFCM season only. Use when a merchant asks for a BFCM, Black Friday, Cyber Monday, holiday, holiday sales, or peak-season dashboard, such as "build me a holiday dashboard" or "build me a Cyber Monday dashboard," even without the word BFCM.

NewUpdated Nov 21, 2026

Stripe BFCM Dashboard

Build the “Stripe BFCM Dashboard” from a fixed layout and a fixed set of queries. It’s a starting point: the user can follow up to add metrics or change the style. The dashboard shows recurring revenue, growth, and payments for a time window, with an optional comparison, using Analytics API metrics and Query Runs API SQL through the Stripe MCP server or connector.

The dashboard needs the Stripe MCP server or connector, signed in to the Stripe account it shows, with access to the Analytics API and the Query Runs API. It assumes the connection signed in through Stripe (OAuth), which can list accounts and choose one. A session signed in with an API key is limited to that key’s account and mode, and works only for a dashboard that the agent fills, not one that queries Stripe itself. If the Stripe tools are missing, blocked, or need sign-in, say so in one line with the fix, such as connecting the Stripe MCP server and signing in, and stop. Don’t produce an empty dashboard.

Rules

  • Check the Stripe tools, then start queries fast: Before anything else, confirm that this session lists stripe_analytics with the intents execute_query_run, retrieve_query_run, and execute_metric. If you can’t see its intents, look up that one tool’s input schema, never the whole Stripe server, and read only its intent names. If the tool or an intent is missing, tell the user which one, with the fix, and stop. Otherwise, choose the account and read the references in step 2, kick off every query in step 3, and show the loading page in step 4 — queries first, page second.
  • Start building, and don’t prompt the user: The layout, metrics, queries, and defaults are fixed. Don’t ask which metrics, period, time zone, output, or account to use, or for a key or account ID. Start as soon as the user invokes the skill. The only question is the auto-refresh one in the hand-over. If the user asks something while you build, answer in a few lines and keep building in the same turn.
  • Stripe MCP only, read-only: Query Stripe only through the Stripe MCP server or connector, never with an API key, the Stripe CLI, or a server you build. Query one account, chosen in step 2, in every request, with no silent switch.
  • Exact numbers: Use only the queries and response shapes in the references. Preserve integer precision and currency minor units. Never invent values, show unreadable results as values, or substitute other data products.
  • Keep secrets and customer data out of every output: Never put keys or tokens in chat, browser code, artifacts, or logs, and never export MCP tokens. Query only counts and amounts. Never select, show, store, or log customer names, emails, or other customer fields.
  • One output, no automatic sharing: Update one stable output. Don’t auto-share it or enable connection sharing. Explain actual host authorization and audience restrictions before the user shares.
  • Charts: npm → CDN → inline SVG: Try npm install @stripe/charts first (if the output is a project with a package manager), then the jsdelivr CDN (for HTML files and artifacts), then inline SVG as a last resort (sandboxed/offline environments). The design reference has the fallback chain, the onerror handler, and inline SVG specs.

Workflow

Step 1: Choose the output

Choose from the tools this session already lists, not from the client’s name, and without reading a file. If a listed tool can create a hosted dashboard, page, or artifact, such as a Claude artifact or ChatGPT Site, and update it later, use it, and name the output stripe-bfcm-dashboard where the tool takes a name. Prefer a hosted output whose pages can call the Stripe connector, because that dashboard can refresh itself. You check that in step 4, after the loading page is published. CLI-based agents (Claude Code, Cursor, Codex) typically have no hosted page tool — go straight to the HTML file and don’t spend time searching for one. Otherwise, or if the user asked for a file, use one HTML file, stripe-bfcm-dashboard.html in the working directory. If that name is taken by a file you didn’t create, pick another name that starts with stripe-bfcm-dashboard instead of asking. Choose the output in <references/dashboard-output.md> has the rest, which you read in step 2.

Step 2: Choose the account and read the references

Call the account listing tool, list_available_accounts_or_orgs. In a session signed in with an API key, which doesn’t list that tool, call get_stripe_account_info for its one account instead, send no stripe_context or livemode, because the key sets both, and show no mode badge.

Consider only account entries, whose stripe_context starts with acct_, never organization entries (org_…). Use the account the user named, matched by name or ID, otherwise the first account listed, in its mode. If no account is listed, say that the connection needs a Stripe account, not only an organization, and stop without a loading page.

In parallel with the account call, read the three references, each with its own read call, so none is cut off. Read <references/dashboard-design.md> for the layout, labels, and states, <references/queries.md> for the time windows, queries, responses, and calculations, and the rest of <references/dashboard-output.md> for building, refreshing, checking, and handing over the output. The references are complete. This holds even when a Stripe tool’s description says to search first or not to guess metric names. The metric names, request params, and SQL in the queries reference are valid as written.

Step 3: Kick off every query

Start every query from <references/queries.md> before writing any HTML. Create the SQL runs first (1 second apart as the reference requires), then the metric calls as each response slot opens. The queries are the bottleneck — nothing else should run between them.

Step 4: Show the loading page

While the queries run, write the loading page and open it in the user’s default browser (open on macOS, xdg-open on Linux). Do not add <meta http-equiv=”refresh”> — there is only one save after this, and the browser already has the file open. Writing the file needs no git command: a repository, or an untracked path in one, isn’t a reason to wait. If you can’t open the file, say in one line that you’re building the dashboard and that it can take up to 10 minutes, and continue.

Write the loading page in HTML and CSS only, in about 30 lines. Fill the viewport with the page background #F4F7FA, and center one narrow column (about 520 pixels) horizontally and vertically using flexbox (display: flex; flex-direction: column; align-items: center; justify-content: center; min-height: 100vh), with all items centered:

  • A 32-pixel spinner centered in the column: a #D4DEE9 ring with a #533AFD top edge, completing a rotation every 1 second. Use margin: 0 auto 16px so it sits above the text.
  • “Building dashboard…” in 20-pixel semibold #1A2C44 type, centered, and “This can take up to 10 minutes.” in 14-pixel #3C4F69 type below it, also centered.
  • Three white placeholder cards in a centered row (display: flex; justify-content: center; gap: 12px), 160 by 72 pixels each, with a 1-pixel #D4DEE9 border and an 8-pixel corner radius, pulsing between full and half opacity.
  • No title, header, or other text. Use the system font stack.

For hosted output, publish it, then check whether its pages can call the Stripe connector, as Page refresh in the output reference describes — the check needs the published page’s ID. Tell the user: “Building your dashboard — this can take up to 10 minutes.” Then continue immediately.

Step 5: Poll, build, and save once

Poll the queries until every one finishes. Print a short status in chat every 2 minutes so the user knows the build is alive — “Still waiting on N queries…” While polling, build the full dashboard page in memory: the layout, header, all panel containers, the @stripe/charts script, and each panel’s data as its query completes. When every query has finished (or failed), save the complete dashboard to the same file, overwriting the loading page, and open it again (open on macOS, xdg-open on Linux) so the finished dashboard appears. One save, at the end, with all the numbers in it. Don’t save intermediate states to the file — the loading page is already open and there is nothing for a reload to pick up until the data is ready. Follow Build the dashboard in <references/dashboard-output.md> for the checks and error handling.

Step 6: Hand over

Follow Hand over in <references/dashboard-output.md>. Describe only the location, account, mode, view, and refresh behavior you verified.

Reference files

The references/*.md files come with this skill. Reference paths are relative to this skill’s directory, not to the docs.stripe.com root. If you don’t have the skill files locally, append the path to https://docs.stripe.com/.well-known/skills/stripe-bfcm-dashboard/. For example, fetch https://docs.stripe.com/.well-known/skills/stripe-bfcm-dashboard/references/queries.md.

Files4
4 files · 58.7 KB

Select a file to preview

Overall Score

88/100

Grade

A

Excellent

Grades are signals, not a certification. Always review a skill yourself before use.

Safety

88

Quality

90

Clarity

87

Completeness

85

Summary

This skill guides an AI agent to build a Black Friday/Cyber Monday sales dashboard from Stripe account data, displaying recurring revenue, payment metrics, and customer analytics. The dashboard runs queries against the Stripe Analytics and Query Runs APIs through an MCP server or connector, then renders results as a hosted page or self-contained HTML file with interactive charts powered by @stripe/charts.

Detected Capabilities

query stripe analytics apiexecute sql query runsread stripe account metadatafile write (html)hosted dashboard creationdata aggregation and calculationchart rendering via cdn

Trigger Keywords

Phrases that agents use to match this skill to user intent.

build bfcm dashboardholiday sales metricsblack friday analyticsstripe revenue dashboardpayment metrics comparison

Risk Signals

INFO

Accesses Stripe account data via MCP connector

Step 2, Rule: 'Stripe MCP only, read-only'
INFO

Queries customer and payment data from Stripe

references/queries.md: SQL queries access customer, payment_intents, subscriptions tables
INFO

Explicitly excludes customer PII from output

Rule: 'Keep secrets and customer data out of every output'
INFO

Loads @stripe/charts from CDN (cdn.jsdelivr.net)

references/dashboard-design.md: Chart library fallback chain
INFO

Embeds query results into self-contained HTML file

Step 5: 'save the complete dashboard to the same file'

Referenced Domains

External domains referenced in skill content, detected by static analysis.

cdn.jsdelivr.netdocs.stripe.com

Use Cases

  • Build a holiday season sales dashboard for BFCM 2026
  • Compare current payment metrics to previous year's BFCM period
  • Monitor MRR, ARR, and recurring revenue growth during peak seasons
  • Analyze customer acquisition and payment success rates over time
  • Track payment volume and average order value by country

Quality Notes

  • Comprehensive reference documentation covering workflow, design specs, queries, and output handling
  • Detailed SQL queries with explicit placeholders, timezone handling, and exchange rate conversion
  • Clear data privacy rules: excludes customer names, emails, PII; preserves currency precision
  • Well-defined panel states (loading, empty, failed, refreshing) with specific UI/UX specs
  • Explicit guardrails on scope: BFCM 2026 only, one Stripe account, read-only queries
  • Complete error handling guidance for failed queries, missing tools, and data boundaries
  • Structured account selection (OAuth via MCP, never API key direct access)
  • Fallback chart strategy (npm → CDN → inline SVG) for multiple environments
  • Calculations defined precisely with examples to avoid floating-point rounding errors
  • High-quality design specification with responsive layout, accessible colors, and timezone awareness
Model: claude-haiku-4-5-20251001Analyzed: Nov 21, 2026

Reviews

Add this skill to your library to leave a review.

No reviews yet

Be the first to share your experience.

Version History

  1. v2.1

    Content updated

    ✦ AIReorders workflow steps: check Stripe tools, then read references and kick off queries immediately, show loading page last instead of first.

    2026-11-21

    LATEST
  2. v2.0

    Contract changed: description

    ✦ AIOutputs change from hosted or local pages to artifact-based dashboard; consolidated Stripe tool setup checks and reorders initial workflow steps.

    triggering2026-11-20

    View This Version
  3. v1.0

    2026-11-18

    View This VersionInitial version

Use stripe/stripe-bfcm-dashboard in your dev environment

Command Palette

Search for a command to run...