Inspiration
In Korea, sending a coffee or cake voucher through a chat app is a common way to say thanks, and shops send e-gift cards and coupons too. We save them as screenshots, they sink into the photo library, and it is easy to find one only after it has expired. The phone photo backup already sits on the NAS, so the NAS is the one place that can look back through all of those pictures.
What it does
Voucher Rescue is a Skill for UGREEN AI NAS. One request, such as "Find my vouchers", starts a guided run over the photo dates and libraries I choose. The NAS photo search finds images whose recognized text has voucher words in Korean or English. The NAS vision model reads each one: brand, item, expiry date, used stamp, and only the last four digits of the code. The Skill sorts the vouchers into expiring soon, valid, expired, used and unsure. I review the summary, mark the ones I already used, and approve. Then it can create a "Use this week" album to open at the counter, tag the files, and save a plain list, so a later question like "Do I have a coffee voucher?" gets a quick answer. Photos are never moved or edited, full codes stay in the images, and nothing leaves the NAS.
How I built it
With the official Uliya Skill Kit 0.7.3. The Skill is Markdown only: SKILL.md delegates each step to registered UGOS experts (photo-agent, file-agent, knowledge-agent) and handles every delegation state (completed, failed, cancelled, waiting_user, waiting_approval). Six reference files hold the reading rules, the output format, the privacy rules, the test plan, a user guide and the disclosures. The kit's static checker returns static-pass.
Challenges
Designing without a NAS in Round 1: every step had to map to an action in the capability catalog, every write had to wait for approval, and voucher codes had to stay out of every output.
What's next
On the finalist test day: run a synthetic set of 24 voucher images with an answer key, check OCR on stylized Korean vouchers, and confirm how text filters, albums and tag approvals behave on a real NAS.
Built With
- file-agent
- knowledge-agent
- markdown
- photo-agent
- ugos
- ugreen-ai-nas
- uliya
