Inspiration

Creators rarely lack material. They have too much of it. A library of 2,000 photos and clips leaves them stuck deciding where to start. NAS AI helps you find files, but nothing helps you start.

What it does

Creative Constraint Generator reads a creator's own NAS media library, proposes three deliberate, executable rules (for example: "only files shot 5-7 pm, no faces"), each with a real file count, and, after the user approves, copies only the matching files into a new staging folder with a one-page constraint card explaining why each file qualified. A Brief mode does the same for a client request, turning it into a shortlist. Originals are never modified.

How we built it

A single orchestration SKILL.md for the UGREEN AI NAS. The parent Skill only delegates to four experts (photo-agent, file-agent, document-agent, knowledge-agent), uses verified results only, never invents data, and handles all five delegation states plus empty and partial results. Folder creation, copying and card saving sit behind platform approval. No scripts, external AI or credentials.

Challenges

Exact sunset times are not available, so time rules use clock windows. Rule quality depends on metadata and recognition results, so files lacking them are excluded and counted, never guessed.

Accomplishments

The Skill package passes the official Uliya static check with no errors or warnings, with edge cases and a test plan documented.

What we learned

Designing around real capabilities and approval boundaries made the Skill simpler and more trustworthy.

What's next

Round 1 status: designed and statically checked, not yet tested on a NAS. If selected as a finalist, we will test on the real NAS: two core cases (constraint mode and brief mode) and a zero-match edge case.

Built With

Share this project:

Updates

Submission history