Refund and Returns Desk: agent fixes for your return policy, delivery page and delay notice, policy matched, refund amounts checked, owner approved
by PaperChaseWebb
Let an agent tidy your return policy, delivery page, product page and delay notice, check refunds and late orders against your order list, and review replies to refund requests, without inventing a window, a fee or free returns, or telling a customer a refund was made when it was not. It writes fixed copies and never touches the live site, refunds anything or sends anything. A fix check refuses a removed return window, a deleted sentence, a new promise or a new email. You approve at your terminal and paste it in yourself.
- Price
- $29
- Difficulty
- Intermediate
- Setup
- 45 min
- Format
- .md file
refund-returns-desk-policy-matched-amounts-checked-owner-approved.md · 209 KB · instant download after checkout
Want this built for you instead? We can build it for you.
What you end up with
Your return policy says how many days and from when, how to start a return, how and when refunds are paid and who pays to send it back; your delay notice offers to cancel for a refund; and you know which orders are late and which refunds are overdue, too big or too small, one approved change at a time, with nothing deleted and no claim that anything is compliant.
What you get
- 21 sections and an appendix in one .md file, with a one-line unpack for all 76 files
- returns_check.py pages: 20 checks on the return policy, delivery page, product page and delay notice against your own terms, from a window with no start day and no way to start a return to store credit only, a restocking fee nobody set, "All sales are final" under a window and a delay notice with no cancel option
- returns_check.py orders: 11 checks on your order list, so you see late orders with no notice, cancelled orders not refunded, returns with no refund, refunds too big or too small, final sale refunds and open chargebacks
- returns_check.py requests: 13 checks on replies to refund and return requests, such as a refund claimed but not made, a card number asked for, a refusal inside the window and "please drop the chargeback"
- returns_check.py fix: 12 checks that compare the agent's copy with the original, so a deleted sentence, a removed return window or start link, a new promise, email or fee fails
- A claims check that stops reports saying a page is compliant or that the agent refunded, published or sent anything
- The owner step: approve.py needs a real terminal and part of the file hash, refuses agent sessions and piped input, and checks the original has not changed before you mark it published
- An eval set for a made-up running store with 4 pages, an order list and 12 request threads, rules and two skills for Grok Bot, Claude Code, Codex and Hermes, six prompts and five templates
- The test harness: 73 unit tests and 508 planted faults you can repeat
Who it is for
Small online stores and agencies who want an agent to tidy the return pages and watch refunds and late orders, with a person approving every change, every refund and every reply.
Works with
Solid chips: the agent is named in this listing's own tools list. 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: The kit is the only purchase. Everything runs locally with Python; the checker makes no network calls, logs in to nothing and changes nothing on your site or in your store.
Time to result: The worked example in 5 minutes; your own pages and order list checked in about an hour
How it works
Unpack and test
Save the kit as pack.md in an empty folder, run the unpack line from Appendix A, then sh test_all.sh.
Expected: 76 files written; 73 tests OK; 508 of 508 planted faults caught.
Watch the flow
Run sh example/run_example.sh and read it next to section 17.
Expected: A candle store's pages with 20 findings, an order list with 10, the request threads with 7, the agent's first-try fixes caught with 10, the second tries clean, a first-try report with 5 findings and a piped approval refused.
Put in your store
Fill in site/settings.json and site/facts.md from the templates, paste your four pages into site/pages/, export orders to site/orders.csv, then run python3 desk/returns_check.py site and python3 desk/returns_check.py pages.
Expected: A short list of settings and page lines to fix first.
Fix and list refunds
Have your agent follow skills/policy-fix and skills/request-review, then run returns_check.py fix queue/fixes and approve each file at your terminal.
Expected: Fixed files in queue/fixes with 0 high findings, and a list of late orders and refunds for you to handle.
Sample prompts, free to use
Follow skills/policy-fix for every page in site/pages. Do not change the live site. List the codes fixed and the codes left for me, with the reason for each.
Prompt 1 of 6.
Follow skills/request-review for site/orders.csv. Refund nothing and send nothing. List each order id that needs a delay notice or a refund, the amount and the last day, starting with the oldest.
Prompt 2 of 6.
Run returns_check.py pages and requests. For each P10, P19, C07 and C08, tell me the fee, term or promise so I can add it to facts.md or take it off the page.
Prompt 4 of 6.
Free preview of the file
# Refund and Returns Desk: agent fixes for your return policy, delivery page and delay notice, policy matched, refund amounts checked, owner approved Version 1.0, 2026-11-01. By PaperChaseWebb. For small online stores (candles, clothing, gear, beauty, food, home goods, print on demand) and the agencies that run their sites, who want an AI agent to tidy the return policy, delivery page, product pages and delay notice, check refunds against the order list and review replies to refund requests, without inventing a window, a fee or a refund, or telling a customer their money is on the way when it is not. Works with Grok Bot, Claude Code, Codex and Hermes. Python 3.8 or newer, standard library only, no API keys, no logins. The kit and the agent never change your live site or store, never issue a refund and never send anything. Tested with 73 unit tests and 508 planted faults on Linux with Python 3.13, 3.12, 3.11, 3.9 and 3.8. All business, contact and order data in this kit is made up. Not legal advice. Returns are where small stores lose money and trust. The FTC's Mail, Internet, or Telephone Order Rule asks you to have a real basis for the delivery time you state (30 days when you state none), to tell the buyer when an order will be late and offer to cancel for a prompt refund, and to refund a cancelled card order within 7 working days. California and New York ask stores to post their refund policy where buyers can see it; with no posted policy, a buyer can return for a refund
Test record
- Date
- Nov 1, 2026
- What was run
- tests/test_lib.py (8), test_site.py (4), test_pages.py (13), test_orders.py (11), test_requests.py (11), test_fix.py (8), test_modes.py (7), test_mini.py (3) and test_approve.py (8, on a real pseudo terminal); tests/mutate.py with 489 planted faults across pages, order lists, request threads, bad fixes, reports, settings, agent rules and the eval set, plus 19 against returns_mini.py; returns_check.py set on the eval set; example/run_example.sh end to end; and the unpack line with test_all.sh. Linux, Python 3.13, 3.12, 3.11, 3.9 and 3.8.
- Result
- 73 passed, 0 failed on all five versions; 508 of 508 planted faults caught and all clean baselines clean; the eval set matched its 40 expected codes and every fixed page had 0 high findings; the example caught the first-try fixes with 10 findings and the first-try report with 5, and refused piped approval; the unpack line wrote every file and test_all.sh passed there.
- Not run
- A live Grok Bot, Claude Code, Codex or Hermes session. Store and payment settings and live sites. State laws beyond the California and New York posting rules, and laws outside the US. Languages other than English. macOS or Windows. Legal review.
Copied from this item's dated test notes. What was not run stays in.
Worth knowing
- The agent rules were checked by the checker but not loaded inside a live Grok Bot, Claude Code, Codex or Hermes session for this release. The eval fixes were written by hand to the rules. Give your agent the entrance exam in section 15 before you rely on it.
- Not legal advice. Nothing in the kit says a page, a reply or a business meets any law. The checks follow the FTC mail and online order rule and the California and New York refund policy posting rules; other states and countries were not checked. Section 20 lists the public sources it was built from.
- The checker reads the text you copy out of your site and the list you export. It does not read your store or payment settings or see what the live site shows today, so open the policy on a phone and start a return yourself after each change. English only.
- The owner step assumes the agent cannot run approve.py at a terminal. Run it as a separate user; agent rules alone are not a hard boundary.
- All businesses, orders, names and numbers in the kit are made up. Tested on Linux only; macOS and Windows were not tested.
- On 2026-11-01 on Linux: 73 unit tests and 508 of 508 planted faults passed on Python 3.13, 3.12, 3.11, 3.9 and 3.8; the eval set matched its 40 expected codes; the unpack line wrote every file and test_all.sh passed in that folder.
Questions
Which agents does it work with?
Any agent that can read files and run a script. The kit has rules for AGENTS.md and CLAUDE.md, Claude Code and Codex notes, Hermes notes and a Grok Bot standing order. Those were checked by the checker, not loaded in each agent for this release.
Will it change my live site, refund orders or reply to customers?
No. The agent writes fixed copies into a queue folder and a list of orders that need a notice or a refund. You approve each file at a terminal, paste it in yourself, and issue refunds and send replies from your own store and email.
Which stores does it work with?
Any store that lets you copy your page text and export orders to CSV. The kit never logs in to the store.
Can I keep a restocking fee or final sale items?
Yes, if they are in your settings. The checker then makes sure the page states the fee with its amount and names each final sale item, and flags a fee or a blanket "All sales are final" that your settings do not back.
Can you set it up for us?
Yes. Our Install sprint sets up your settings, facts file, order export and eval set, or Managed Ops runs the fixes and refund lists for you each week. Book a call or write to [email protected].
Details
Version 1.0 · Updated Nov 1, 2026 · Markdown version of this page
License: Use inside your own business or for your clients. Do not resell or republish the kit itself. Download terms
Related
Refund and Returns Desk: safe return policy rules and a ten-check policy scan (free sample)
From Refund and Returns Desk: the ways return policies slip, ten rules to paste into your agent, and a read-only script that runs ten checks on one return policy, from a window with no start day and no way to start a return to no refund method, nothing on return postage, a restocking fee with no amount and "All sales are final" under a window.
Subscription Renewal Desk: agent fixes for your checkout, terms and cancel page, plans matched, cancel path kept, owner approved
Let an agent tidy your subscription checkout, terms, cancel page and renewal reminder, plan the renewal, trial and price change notices from your subscriber list, and review replies to cancel requests, without dropping the renewal line, hiding the cancel path or making up prices and offers. It writes fixed copies and never touches the live site or sends anything. A fix check refuses a removed cancel link, a deleted sentence, a new discount or a new phone number. You approve at your terminal and paste it in yourself.
Review Reply Desk: AI-drafted review replies, checked for privacy and offers, approved by the owner
Let an agent draft your Google and Yelp review replies without the mistakes that cost money. A checker refuses drafts that use a last name or details from your records, offer a discount for a better review, ask for a review to be changed, argue, invent numbers or promises, or repeat your last reply word for word. Health topics, allergies and legal threats go to the owner, clinics get a stricter privacy mode, and nothing is posted until the owner approves it at a real terminal.
Email Sender Fix Desk: agent fixes for one-click unsubscribe, footers and SPF, DKIM and DMARC, checked against the original, owner approved
Let an agent put the one-click unsubscribe headers, a visible unsubscribe link and your postal address into your newsletters, and propose clean SPF, DKIM and DMARC records, the things Gmail, Yahoo and Outlook ask senders for. It writes fixed copies and never sends or touches DNS. A fix check refuses deleted words, made-up addresses, tracking pixels and hidden links. A records check refuses dropped senders, invented keys and DMARC jumped straight to reject. You approve at your terminal and make the change yourself.
Invoice Chase Desk: AI-drafted payment reminders, checked against your ledger, approved by the owner
Let an agent draft your overdue invoice reminders without chasing a client who already paid. A planner reads your ledger and decides who gets a reminder today, and at which step. A checker refuses drafts with the wrong balance or dates, a fee your terms do not include, threats or pressure, new bank details, lookalike payment links, other clients' invoices or claims nobody can back up. Disputes, money trouble and odd replies go to the owner, and before anything is recorded as sent the desk checks the ledger again.
Don't Get Sued: a pre-launch legal-risk gate for AI-built apps and agents
A read-only agent gate for the fourteen legal risks that hit small builders most: marketing texts and email, trackers before consent, accessibility, kids' data, biometrics, plain-text personal data, AI disclosure, AI claims you cannot prove, takedowns, policies that do not match the product, renewals and refunds, licenses, and agents that act for users. Evidence rule, gate register, seeded test app, 12 prompts, counsel packet. Not legal advice.
Ask-first rules for AI agents: hard nos, approval cards and a test set
Paste-in rules that make an agent ask before it sends, spends, posts or deletes, define what a real yes is, and stop retries after a no, plus 12 test scenarios to check it.