Client Project Wrap-up

Inspiration

Photographers, editors, designers and freelancers keep their client work on a NAS, and every project ends with the same chores: gather the final files, check what the contract and invoice say, prepare the delivery, and write it all up. It is repetitive, easy to get wrong, and easy to put off. The hackathon's "Automate Your Workflow" track asks for exactly this kind of repetitive work to become an executable AI workflow, so I wanted to turn project close-out into one sentence: "wrap up the Ahmed wedding project."

What it does

Client Project Wrap-up is an AI Skill for UGREEN AI NAS. From one request it:

1. Finds the project's photos, videos and documents.
2. Reads contracts and invoices and pulls out the parties, amounts and due dates.
3. Shows a plan and waits for the user's confirmation.
4. Copies the final files into a delivery folder and creates a delivery album.
5. Saves a short project summary next to the delivery files.

It chains four NAS capabilities (files, documents, albums and knowledge) into one finished result, and client data stays on the local NAS.

How we built it

I built it with the official UGOS Skill Creation Tool and Claude. A Skill is an orchestration workflow written in SKILL.md: it delegates each step to the NAS's built-in experts (file, document, photo and knowledge), and every step uses only the real results of the previous one. I checked each step against the capability catalog in the tool, so the Skill only uses capabilities that actually exist. I then ran UGREEN's static checker until it reported 0 errors and 0 warnings.

Safety is built into the design:

• Original files are never moved, renamed or deleted. Files are only copied.
• Nothing changes until the user confirms.
• Existing files are never overwritten.
• If a search returns no results, the Skill says so and stops instead of guessing.
• Every delegated step has explicit handling for completed, failed, cancelled, waiting for the user, and waiting for approval.

Challenges we ran into

This is my first hackathon, and I don't own a UGREEN NAS, so I could not run the Skill on real hardware. The biggest challenge was designing a workflow that is correct on paper. I had to stay strictly inside the documented capabilities and learn the checker's rules, for example that only SKILL.md and Markdown references are allowed in the package and that every step must handle empty and partial results. The part I am least sure about is how photos get added to the delivery album, so I will test that first on the real NAS.

Accomplishments that we're proud of

• A complete six-step workflow that chains files, documents, albums and knowledge in one request.
• A Skill that passes UGREEN's static checker with 0 errors and 0 warnings.
• A safe-by-default design: confirmation before any change, copy instead of move, and no guessing when nothing is found.
• Finishing a first hackathon submission with no prior NAS experience.

What we learned

How a Skill delegates work to NAS experts, why confirmation and empty-result handling matter for anything that changes a user's files, and how much a clear, narrow idea helps. I also learned to design from the catalog of what really exists instead of from what I imagined.

What's next for Client Project Wrap-up

If I am selected as a finalist, I will deploy and test the Skill on the provided UGREEN AI NAS iDX6011 Pro, fix whatever the real environment reveals, and record the final demo there. After that I want to support several projects at once and add a client-ready delivery note.

Built With

• claude
• markdown
• ugos
• ugos-skill-creation-tool
• ugreen-ai-nas
• vibecoding