When should you use it?
- You chase refunds now and then and want a weekly habit instead of a yearly cleanup
- Order, receipt and renewal emails land in one inbox the agent can read
- You want one number each week with the detail behind it
Not for: A one-time deep pass over a year of statements (use statement-subscription-cancel-audit); Letting the routine cancel, return or dispute anything without asking you; Counting an unexplained charge as owed before you have looked into it
What do you need, and what do you get?
You need
- Read access to the inbox where orders, receipts and renewal notices arrive
- Your running refunds table, or permission to start one
- The day and time for the check, and where the summary should go
You get
- New rows each week for late orders, charges with no matching receipt, and renewals due within a month
- A total split into confirmed money owed and items still to check
- A short Friday message with that total and the three most urgent follow-ups
How does it work, step by step?
- Pick the inbox and tableGrant read access to the inbox and point the agent at your refunds table, or let it create one with columns for status and evidence.
- Define the three checksOrders past their promised delivery date, card or bank charges with no receipt to match, and renewal notices for the month ahead.
- Split owed from possibleCount money as owed only when a refund is confirmed or an order clearly failed. List everything else as to check.
- Schedule itSave it as a Friday routine with a fixed time and a short message format, so the report reads the same each week.
- Act yourselfUse the report to choose follow-ups. Returns, cancels and disputes stay with you or a separate approved playbook.
The prompt
The original prompt by @free_ai_guides, word for word, with a link to where it was published. Below it is our adapted version with fill-in blanks, if you want it. Works in Grok Bot, ChatGPT agent and Claude. Values in lime are placeholders.
Original prompt by @free_ai_guides
Every Friday, check my inbox for late deliveries, charges that don't match a receipt, and renewals coming in the next 30 days. Add them to the table and message me the total I'm owed this month. Save this as a routine.
Exactly as published by @free_ai_guides in X thread: 7 prompts to get your money back (prompt 7, the Friday check).
Our adapted version (optional)
Fill in the blanks (optional; the copied prompt uses your values)
Set up a weekly read-only money check. Inbox: {inbox}. Refunds table: {table_location}. Run time: {day_and_time}. Send the summary to: {delivery_channel}. Each run: 1) Find orders past their promised delivery date, card or bank charges with no matching receipt, and renewal notices due within 30 days. 2) Add new items to the table with date, merchant, amount, evidence link and status (owed / to check / upcoming renewal / closed). Do not duplicate existing rows. 3) Count as owed only refunds the merchant confirmed or orders that clearly failed. Everything uncertain is to check. Renewals never count toward the total. 4) Message me: confirmed owed this month, amount to check, renewals coming up with dates, and the three most urgent follow-ups. 5) Never contact a merchant, cancel, return or dispute anything. Ask me first.
What can go wrong?
- Renewals are upcoming charges, not money owed. Keep them out of the total.
- Receipts often land in another inbox or app, so a charge without one is a question, not proof of an error.
- A stale tracking page is not a lost parcel. Check the seller's promised date before chasing.
- The total drifts upward if nobody closes rows. Mark items received or dropped each week.
Our take
Most money people are owed is lost to forgetting, not to refusal. The thread's last prompt fixes that with a standing weekly sweep: late orders, charges you cannot match to a receipt, and upcoming renewals, all logged to one table with a total sent to you. It works best as a routine because the value comes from catching things while return windows and cancel dates are still open. We changed one thing on purpose. The original sends a single total; we split it into money that is confirmed owed and items that still need checking, and we keep renewals out of the total because they are charges to decide on, not refunds. Without that split, an unmatched charge that turns out to be a forgotten gift card purchase inflates the number and erodes trust in the report. Keep the routine read-only; follow-ups go through playbooks that ask first, such as unwanted-charge-refund-recovery-audit or card-dispute-pack-after-a-refused-refund. In our dry run with a fictional inbox (a $64 order three days past its promised date, a $31.99 charge with no receipt, a $25 refund the seller confirmed but had not paid, a $119 annual renewal due in 15 days, and a normal charge with a matching receipt), the agent logged four rows, reported $25 confirmed owed and $95.99 to check, listed the renewal separately with its date, ignored the matched charge, and changed nothing in any account.
Where the time goes back: Small refunds and return windows expire quietly. A five-minute read on Friday keeps them, and next month's renewals, in view before they pass.
FAQ
How is this different from personal-subscription-audit-routine?
Will it message me when nothing changed?
Does it chase merchants for me?
Source and credit
The original prompt on this page is by @free_ai_guides, shown word for word with a link to where it was published. The summary, notes and adapted version are ours.
- Original
- X thread: 7 prompts to get your money back (prompt 7, the Friday check)
- Creator
- AI Guides (@free_ai_guides)
- Published on
- X
- Original date
- Oct 25, 2026
- License
- Not openly licensed; the original prompt is shown word for word with credit and a link to the original
- Originally for
- Grok Bot
- AI dry run
- Oct 26, 2026: passed. Dry-run tested by our AI test agent on sample input. Owed vs to-check split held; renewals kept out of total; read-only.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails. - Verified
- Oct 26, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.
Are you the creator? Claim or correct this page.