What They Know

See what companies hold about you, without handing it to another cloud: a UGREEN NAS Skill that audits your personal-data exports locally and builds a masked report.

Comment

Inspiration

Under GDPR anyone can ask a company for a copy of their data, and services like Google Takeout let you download it. The copies arrive as archives of JSON, CSV, HTML and PDF that few people ever read, yet they are among the most sensitive files a person owns: some contain AI-written customer profiles, some arrive locked with the customer's national ID number as the password. Today the easy way to make sense of them is to upload them to a cloud AI, which defeats the point.

What it does

What They Know is a UGREEN AI NAS Skill. Point it at a folder of exports (one subfolder per company) and it:

  • inventories each company's files and lists archives that still need extracting,
  • reads what each company holds, by category: identity, contact, government ID, payment, location, purchases, messages, device and logs, profiling, third-party data,
  • finds sensitive values in text and images and shows them masked (ID, card and account numbers by their last 4 digits; everything else by type only),
  • builds a timeline and a per-company action list: keep, ask for erasure (GDPR Art. 17), or ask for what the export visibly left out (Art. 15),
  • on request, saves the report, tags the folders (with undo) and drafts request letters. It never sends anything and never contacts a company.

How it is built

An orchestration Skill written with the official UGOS Skill Creation Tool. Each step is delegated to Uliya's experts (file-agent, document-agent, knowledge-agent, photo-agent), and only real results are passed forward. It is read-only by default; saving and tagging go through approval; all five delegation states are handled, and empty results are reported, never filled. The official checker reports static-pass with 0 warnings.

Challenges

There is no archive-extraction capability and no pattern search for ID numbers, so the Skill is explicit about coverage ("read N of M files") and treats detection as best effort.

What's next

Testing on a real UGREEN AI NAS with a synthetic set of fake company exports containing ten planted specimen numbers.

Built With

Share this project:

Updates

Submission history