Inspiration With the explosive growth of personal data, smart households and small businesses are increasingly turning to Network Attached Storage (NAS) to keep their files private, secure, and under their own control. However, organizing hundreds of unmanaged files—ranging from tax documents and legal contracts to utility bills and personal IDs—remains a tedious chore.
Worse yet, mainstream AI document organizers rely on external cloud APIs, posing severe privacy risks when handling sensitive financial and legal records. We were inspired to build VaultGuard to bridge this gap: bringing intelligent document triage, sensitivity classification, and PII-redacted auditing directly to local hardware using UGREEN AI NAS and UGOS.
What it learns / How we built it VaultGuard was engineered following an agentic orchestrator-worker pattern tailored specifically for the Uliya Skill Development Standard.
Local Parsing & OCR: We designed the Skill to leverage native UGOS document extraction and OCR capabilities to parse files in place without external transmission.
Sensitivity Taxonomy: Using structured Markdown rules, the system evaluates document content, assigns sensitivity tiers (HIGH vs. STANDARD), and identifies sensitive entities.
Collision-Safe Routing: The orchestration workflow standardizes filenames into a clean, searchable format (YYYY-MM-DD_Category_Issuer.ext) and routes files to permission-isolated vault directories while generating a secure Markdown audit log.
Challenges we faced The primary challenge during Round 1 was designing a robust, fully autonomous agent workflow without requiring physical access to a NAS during initial conception. We solved this by separating the logical orchestration—defining explicit input validation, error-handling fallbacks for corrupted or password-protected files, and permission checks—into clean, modular Markdown specifications (SKILL.md and reference rules) that map directly onto UGOS capabilities.
What we learned We gained a deep appreciation for edge-first AI architecture. Designing VaultGuard reinforced the reality that users want the power of modern AI assistants (summarization, categorization, and entity extraction) without sacrificing data sovereignty or exposing private financial and legal documents to third-party cloud servers.
