-
This image is currently processing. Please be patient...
What it does
After a fire, a flood or a break-in, an insurer's first question is what you owned and what it was worth. Almost nobody can answer it. The proof exists, but it is scattered across invoice PDFs, scanned receipts and thousands of phone photos.
This Skill builds the answer from what is already on the UGREEN AI NAS:
- It reads the receipts and invoices in the folders you name and extracts every item, store, purchase date, price and warranty.
- It finds receipts you only ever photographed and reads them with vision OCR.
- It searches your photo library by content for pictures of each valuable item, so each item has photo evidence of ownership as well as a receipt.
- It saves one dated inventory on the NAS: every item with its value, warranty end date, and links to the receipt and photos that prove it, plus a total and a list of what still needs proof.
- With your approval, it tags every proof file "home-inventory" so the evidence can be found in seconds after a loss.
What problem it solves
Underinsurance and slow, disputed claims both come from the same gap: households can't prove what they owned. Building a home inventory by hand takes days, so almost nobody does it. The documents and photos are already on the NAS, but nobody has the time to connect them. This Skill does it in one request and refreshes it whenever you ask.
How it works
The Skill orchestrates four of the NAS's own experts, passing only real results downstream:
document-agent(ug-document extract, receipt and invoice schemas, deduplication on): items, prices, dates and warranties from documents.photo-agent(ug-photo search-screen+vision-ocr): reads photographed receipts; then searches photos by content for each valuable item.knowledge-agent(ug-knowledge save-text): saves the dated inventory as Markdown on the NAS, with the user's workflow approval.file-agent(ug-filemgr tag-add/tag-add-file): optional evidence tagging, with gateway approval.
Guarantees: it never estimates a price, invents a date or attaches a photo the photo expert didn't return for that item. Every item links to its proof. Nothing is moved, renamed or deleted. Empty and partial results are reported as they are, and approval is required before anything is saved or tagged.
What makes it different
- It connects data that lives in different places. Most NAS skills work inside one library. This one joins documents and photos into a single piece of evidence per item.
- It suits a NAS particularly well. Receipts and home photos are among the most private records a family has. Here they are processed on the household's own device, and the inventory stays there.
- It's ready before the bad day. It turns a scattered archive into a document you can hand an insurer, refreshed whenever you buy something.
Use of AI tools (disclosure)
Built with Claude (Anthropic) via Claude Code, which designed the workflow against UGREEN's capability catalog and wrote the Skill. Validated with UGREEN's official checker. The full disclosure is in references/disclosure.md inside the package.
Built With
- claude
- claude-code
- markdown
- ugos
