What it does
Promise vs Reality is a Skill for UGREEN AI NAS that checks whether the contract, invoices and receipts of one project stored on the NAS agree with each other. It shows every difference with its sources (file and page) and, only after the user approves, saves an evidence report as a new note.
Why
Households and small practices keep the paperwork of a renovation, repair or purchase. Nobody checks whether what was billed matches what was agreed until money, a tax deduction or an insurance claim depends on it. These papers are private and should stay at home, which is exactly what a local AI NAS is for.
How it works
- The user names a project folder and a report folder.
- The Skill lists the documents and reads them: contract amount and payment schedule, invoice totals, receipts.
- It reconciles contract and invoice amounts with the NAS deterministic reconciliation.
- Optionally, it finds photos of the works period.
- A report separates confirmed facts, differences to verify and items not verified, always citing file and page.
- After the user's yes and the platform approval, the report is saved.
A difference is never declared to be an error, and a missing receipt is reported as "not found", never as "not paid". The Skill gives no legal, tax or financial advice.
NAS capabilities used
- ug-filesystem: lists the project documents
- ug-document: structured reading with citations, and contract and invoice reconciliation
- ug-photo: finds photos of the works period
- ug-knowledge: saves the approved report
Status (Round 1)
The pre-test Skill package is SKILL.md plus three reference files, written to the Uliya Skill Development Standard and passing the official static checker. Not yet tested on a NAS: extraction accuracy on real scans, reconciliation with several invoices, report wording.
Limits
Markdown-only orchestration, at most 40 documents per run, reading depends on scan quality, and the comparison covers amounts only. A missing receipt does not prove non-payment, and photos do not prove that work was done.
How it was made
Written with the help of an AI assistant and the official UGREEN Skill Developer Package.
Built With
- ugos
