Skill: Amazon listing writing
Version 1.0 | Last verified 2026-09-17 | Works with Claude Code, Codex, ChatGPT projects, or as plain instructions
Download the skill (zip) - get it on GitHub - or read SKILL.md directly at /skills/amazon-listing-writing/SKILL.md.
What it does
Section titled “What it does”You give it the product facts, meaning materials, measurements, process and test data, plus the marketplace, the category and the keywords you are allowed to use. It produces:
- a feature-to-benefit table, so you can see which claims have a fact behind them and which are only adjectives
- the list of facts you still have to get from whoever knows the product
- finished copy for the title, five bullets, description, A+ modules and backend search terms
- a length and policy check, run by the script in the folder
Who it is for
Section titled “Who it is for”Sellers who write their own listings, and anyone who has asked a chatbot for listing copy and received twenty adjectives and no facts. The position the skill takes is blunt: a claim with no fact behind it gets deleted, not softened.
What is in the folder
Section titled “What is in the folder”SKILL.md the procedurereferences/fab-translation.md feature to benefit, with the questions to askreferences/amazon-fields.md per-field rules, limits, policy red linesreferences/message-metrics.md attention, clarity, effectscripts/check_listing.py length and policy-word checkeragents/openai.yaml display metadata for CodexREADME.md install instructionsInstall
Section titled “Install”The files live in github.com/12222a/amazon-listing-writing, inside amazon-listing-writing-en/, and are copied into the skills folder of whichever assistant you use:
git clone https://github.com/12222a/amazon-listing-writing.git
# Claude Codecp -r amazon-listing-writing/amazon-listing-writing-en ~/.claude/skills/
# Codexcp -r amazon-listing-writing/amazon-listing-writing-en ~/.codex/skills/On Windows, copy the folder into %USERPROFILE%\.claude\skills\ or %USERPROFILE%\.codex\skills\.
Claude Code: copy the folder into ~/.claude/skills/amazon-listing-writing/, or into .claude/skills/ inside a project.
Codex: copy the folder into ~/.codex/skills/amazon-listing-writing/. It is picked up automatically.
ChatGPT or Claude in a browser: create a Project, paste the contents of SKILL.md into its instructions, and attach the reference files to the project.
Anything else: paste SKILL.md as a system prompt.
Checking the result
Section titled “Checking the result”The folder ships with the checker the skill calls:
python scripts/check_listing.py listing.mdOne field: value pair per line. It reports the character count for each field, and every line that hits a policy red line: promotion language, absolutes, medical claims, links, emoji, all-caps words. Lengths are counted in UTF-16 code units, which is how Amazon counts them, so emoji cost two.
Character limits vary by marketplace and category, and they change. The limits in the reference file are common ranges, not guarantees. Confirm in Seller Central before you finalise.
What it refuses to do
Section titled “What it refuses to do”It does not invent specifications, certifications, test data, sales figures or reviews. When facts are missing it names them and stops. It will not use competitor brand names, absolute claims or promotional language, because those get listings taken down in most categories.
It also does not do keyword research, product research or advertising. Those are inputs, not outputs.
Where the method comes from
Section titled “Where the method comes from”The method was distilled from a community article on wearesellers.com (“Do you actually know how to write a listing?”), which is credited inside the skill file. What this package adds: the field-level rules, the order of work, the worked examples, the hard limits, and the checker script that the article does not provide. The wording is written for this skill rather than reproduced from the source.
Version history
Section titled “Version history”- 1.0 (2026-09-17) - first published version.
License
Section titled “License”MIT. Use it commercially, modify it, put it inside your own workflow, and keep the licence text with any copy you redistribute.