Get a weekly metrics report written by an agent from Stripe, analytics and a sheet

What you end up with

Every Monday morning a one-page report is waiting: what moved, by how much, and the two things worth looking at, with every number traceable to its source.

What you get

• Architecture diagram and the data contract between steps
• Collector code for Stripe (restricted key), an analytics adapter and a published sheet
• A deltas module that computes every change in code before the model sees it
• Report prompt that may only cite numbers it was given, with a self-check step
• Delivery options: email draft to yourself, a Slack message or a Markdown file
• Setup guide for read-only keys and secrets

Who it is for

Solo founders and small teams who check five dashboards every Monday and still cannot say in one sentence how last week went.

Works with

Dashed chip: Plain Markdown, so it also pastes into other agents such as Grok Bot, Hermes, OpenClaw or Muse. We list those as compatible by format only, not as tested.

Tools needed

Cost to run: Free tier scheduling on GitHub Actions or Cloudflare Workers. One model call per week on your own key.
Time to result: First report the same day you finish setup

How it works

1. Create read-only keys

   Make a Stripe restricted key with read access only, and an analytics key with the narrowest scope offered.

   Expected: Keys that cannot change anything, stored as secrets.

2. Collect

   Run each collector once by hand and check its JSON against the dashboard.

   Expected: Numbers match what you see in Stripe and analytics.

3. Compute in code

   Run the deltas module to produce this week, last week, change and four-week average for each metric.

   Expected: A table of numbers you could report without any AI.

4. Write the report

   Send only the computed table to the model with the report prompt.

   Expected: A one-page report where every number appears in the input table.

5. Schedule it

   Add the weekly schedule and pick a delivery option.

   Expected: A report arrives at the same time every week.

Sample prompt from the pack

You write a weekly metrics report for [business name]. Input is a JSON table of metrics with this_week, last_week, change_pct and avg_4w for each.
Rules: use only numbers that appear in the input. Do not calculate new numbers. If a metric is missing or null, say so plainly.
Format: a one-sentence headline, then up to five bullets on the biggest moves, then two things worth looking at this week with the reason for each.
Finish with a self-check: list every number you used and the input field it came from.

Free preview of the file

# Get a weekly metrics report written by an agent from Stripe, analytics and a sheet

PaperChase agent use case pack (playbook), version 1.0. Node.js 20, no dependencies. Runs on a GitHub Actions schedule (or a Cloudflare Workers cron with small changes). Setup: 60 to 90 minutes.

## The design in one line

Code collects and computes every number. The model only writes sentences about numbers it was handed, then lists where each one came from.

```
collect (Stripe, analytics, sheet)  ->  metrics.json (raw)
compute (code: this week, last week, change, 4-week average)  ->  table.json
write (model: headline, biggest moves, two things to look at, self-check)  ->  report.md
deliver (email draft to you, Slack webhook or a file)
```

## Why the math stays in code

Language models are unreliable at arithmetic and will produce a confident percentage that is slightly wrong. A report you cannot trust is worse than no report. Every change and average is computed in `compute.mjs`, and the prompt forbids new calculations.

## What you will build

```
metrics-report/
  collect.mjs        # Stripe, analytics and sheet collectors
  compute.mjs        # weekly windows and deltas
  write.mjs          # model call with the report prompt
  run.mjs            # runs all three, saves report.md
  .github/workflows/weekly-report.yml
```

## Contents

1. Keys and secrets (read-only)
2. Collectors
3. Compute
4. Report prompt and writer
5. Schedule and delivery
6. Checking it is right
7. Instructions for your AI coding agent

Worth knowing

• Never give the job a full-access Stripe key. A restricted read-only key limits the damage if a secret leaks.
• Models are bad at arithmetic. The pack computes every change in code for this reason; do not skip that step.
• A report reflects the data sources. If your analytics undercounts because of blockers, the report will too.

Details

Version 1.0 · Updated Oct 15, 2026 · Markdown version of this page
License: Personal and commercial use inside your own business. Do not resell the pack itself. Download terms

Related

Bot template$39

Gmail inbox triage bot: label, summarize and draft replies for review

by PaperChaseWebb

A Google Apps Script bot that labels new mail, drafts replies in your Drafts folder and emails a morning digest to you. Nothing goes to anyone else.

Works withGeminiRuns on your own accounts

Solo business · Intermediate · 25 min setup

Prompt pack$14

Price, pitch and follow up: prompts for solo founders selling services

by PaperChaseWebb

14 prompts for pricing a service, writing a one-page offer, handling objections and following up without sounding desperate.

Works withClaudeChatGPTGeminiPlain .md

Solo business · Beginner · 10 min setup

Agent use case pack$19

Turn a spreadsheet you run by hand into a small web app

by PaperChaseWebb

A staged plan and prompts for turning the spreadsheet your team edits by hand into a web app with logins, forms and an import of your existing rows.

Works withClaude CodeCursorPlain .md

Vibe coding · Intermediate · 20 min setup