What does your Skill do?
PaperTrail Guardian turns documents scattered across a UGREEN AI NAS into a private, queryable household ledger, extracting deadlines, amounts, and obligations, reconciling related documents, and filing everything neatly, entirely on-device.
Who is your target user, and what is their problem?
Families store contracts, insurance policies, invoices, warranties, and receipts on their NAS, but the archive is a black box: renewals lapse unnoticed, duplicate bills slip through, and answering one question means opening many files.
What is your solution?
The Skill scans user-authorized folders, extracts facts (type, parties, amounts, due dates) from documents, links items (contract, invoices, receipts), flags duplicates and mismatches, keeps an obligations ledger, answers plain-language questions from that ledger, and, only after confirmation, tags and files documents into an archive structure.
What is the main user flow?
The user asks to organize documents and report what is due soon. It discovers documents, extracts facts, reconciles them against the ledger, and reports a summary with actions. After confirmation, it files and tags the documents, updates the ledger, and answers follow-ups like 'When does my car insurance renew?'
How is your idea different from existing tools or common AI solutions?
Cloud assistants require uploading sensitive financial and legal documents to third-party servers; chatbots answer one question at a time and change nothing. PaperTrail Guardian is persistent and local: it keeps a ledger, reconciles facts across documents, acts on files in place, and never sends data off the device.
Which UGREEN AI NAS features does your Skill use, and at which step?
Discovery uses Files (browse, find). Extraction and reconciliation use Documents (read, extract information, reconcile) using the local AI model. The ledger and notes are saved via Knowledge (save text). After confirmation, filing uses Files (tag, organize). All processing is local, keeping private data on the NAS.
What is your technical approach?
A coordinator agent orchestrates agents: Discovery (Files), Extraction (Documents), Reconciliation (conflict checks), Filing (Files, confirmation-gated), Ledger (Knowledge); each step consumes the previous step's output. Inputs: authorized folders and a plain-language request. Permissions: read-only by default; writes only to an archive area and the Skill's notes; nothing deleted; confirmation before any move or overwrite. The local model handles classification, extraction, and summaries. No external services, uploads, or credentials are used.
What have you built so far? What is not built yet? What still needs testing on a NAS?
Built so far: the Uliya-standard pre-test package: orchestration workflow, confirmation gates, failure handling, and validation plan, authored with the UGREEN Skill Developer Package. Not yet built: extraction-prompt tuning against real household documents. Needs NAS testing: CLI behavior on a library, extraction accuracy on imperfect scans, and the confirmation flow on the assigned NAS.
How will you test your Skill if you become a Finalist?
Core test 1: a contract, invoice, and receipt in a folder must yield one reconciled ledger entry and a filed set after confirmation. Core test 2: renewal-date and past-payment questions must be answered only from extracted facts. Edge case: a corrupt scan and a duplicate invoice with conflicting amounts must be flagged; nothing changes.
What are the known limits, risks, or outside dependencies?
Extraction quality depends on scan legibility; handwriting is unreliable. Archives need batched runs, as local-model throughput is finite. Date and amount formats vary by country and issuer. The Skill relies only on UGOS capabilities and the local model; the risk is capability behavior differing on real hardware, which Finalist testing will verify.
Third-party components and licenses
None. The Skill is built solely with the free UGREEN Skill Developer Package under the Uliya Skill Development Standard. No third-party code, components, data, or external services are used.
Built With
- ai
- local
- ugos
- ugreen
- uliya
