Inspiration
I kept thinking about how many NAS owners use one folder as a dumping ground for scans, invoices, and contracts. Nothing in there ever gets organized until you actually need something from it, and by then it's too late to find it or catch a renewal date.
What it does
On request, Drop Zone Butler looks through a folder on a UGREEN AI NAS, works out whether each file is an invoice, a contract, or a receipt, and pulls out the vendor, date, amount, and expiry. It shows what it wants to do before touching anything. Once confirmed, it renames and files the document and logs it — so that log quietly becomes a search index and a renewal tracker too.
How I built it
It's a SKILL.md written for UGOS's AI assistant, built entirely on two of UGOS's own supported capabilities — Files (find, organize) and Documents (read, extract information) — rather than inventing anything unsupported. The workflow is six steps: find, read & classify, extract, confirm, organize, log.
Challenges I ran into
My first draft assumed the Skill could run as a background process silently watching a folder. Reading UGOS's own Hackathon Starter Kit guide made clear that Skills are triggered by a request, not an always-on watcher — so I reworked the flow around that, and around requiring confirmation before any file gets touched.
What's next
Testing on a real NAS during the finalist round: how well classification holds up on messy real-world scans, and whether the confirmation step feels natural rather than tedious.
AI tools used
I used Claude to help research the platform's actual capabilities, draft and restructure the SKILL.md and proposal, and catch places where I'd assumed something UGOS doesn't actually support.
Built With
- ai
- claude
- markdown
- nas
- ugos
- ugreen
