Inspiration
Small online sellers who ship from home regularly get complaints like "the box was empty" or "I received the wrong item". The marketplace then asks for proof. The records usually exist, but they are scattered across scanned order sheets, invoices, shipping slips, product photo folders and a packing-station camera. Collecting them by hand for one order is slow and easy to get wrong.
What it does
Given an order number and a complaint type, the Skill finds that order's documents on the NAS, cross-checks the order sheet, invoice and shipping slip, proposes matching product photos and packing-camera events, and, after the owner approves a written plan, copies the confirmed files into one evidence folder with a short timeline summary that masks customer personal data. It reports facts only and never judges who is at fault.
How we built it
The Skill delegates each step to the registered expert agents (file-agent, document-agent, security-monitor-agent, knowledge-agent) and uses only capabilities listed in the Skill Creation Tool catalog 2.0.0. Each step uses the real output of the previous one. The package passes the official checker with 0 errors and 0 warnings.
Challenges
Keeping the workflow inside supported capabilities. The catalog has no camera clip export, so the Skill records camera evidence references instead of copying clips.
Safety
Nothing is deleted, moved or overwritten. Folder creation, copying and saving run only after plan approval. If an order number is found nowhere, the Skill stops and creates nothing.
What's next
If selected as a finalist, we will validate scan extraction, camera queries and the approval flow on a real UGREEN AI NAS.
Built With
- claude
- markdown
- ugreen-ai-nas
- uliya-skill-kit
