Proof of Loss

What it does

Something gets stolen or wrecked and the insurer wants three things: when it happened, proof you owned it, and what you paid. If you have a UGREEN NAS you probably already have all three. The footage is on your cameras, your stuff shows up in the background of your photos, and the receipts are in your documents. Nobody wants to dig through all that the night after a break-in. Proof of Loss does it for you.

Tell it what happened ("my bike and laptop were stolen from the garage last night") and it:

• pulls the camera events from that window
• finds photos of each item from before the loss and asks you to confirm them
• finds the receipts, or a photo of the receipt, and reads the price and date
• builds a claim packet: a timeline, every item marked Strong, Partial or Missing, what's still missing and where to look for it, and a list of every file it used

It never guesses a price, never moves or deletes your files, and never sends anything off the NAS. Albums, folders and saved notes only get made if you say yes.

Inspiration

I run crypto nodes and the monitoring around them, and most of my work comes down to checking whether the evidence actually backs up a claim. Insurance claims are the same problem from the other side. The evidence exists, it's just spread across three places, and the person who needs it is having the worst week of their year.

How I built it

It's a Uliya skill made with UGREEN's skill kit. Each job goes to the expert that owns it: security-monitor-agent for the camera timeline and snapshots, photo-agent for finding the items and reading receipt photos, document-agent for pulling and cross-checking receipt details, and file-agent and knowledge-agent for the optional claim folder and saved note. It passes the kit's checker with 0 errors and 0 warnings.

Challenges

I don't own a UGREEN NAS, so it's built against the published capability catalog, not a real device. The other challenge was keeping it honest. A claim packet that looks complete but has a guessed price in it is worse than nothing, so every step has a rule for empty results, and gaps get listed instead of filled in.

What I learned

Photo search on UGOS goes further than I expected: searching by object, reading text off photos, and describing what's in an image. That's what made the receipt-photo backup possible.

What's next

Testing on a real NAS, then a "before" mode that builds the same packet ahead of time, so the proof is ready before anything happens.

Built With

• ai-agent
• markdown
• ugos
• ugreen-ai-nas
• uliya