Renewal Deadline Radar

A UGREEN AI NAS Skill that finds renewal, expiry, warranty and notice-period deadlines hidden in stored contracts, policies and receipt photos, and lists what to act on first, with evidence.

Comment

Inspiration

Many households and small offices already keep their contracts, insurance policies, warranty cards and receipt photos on a NAS. The problem is that the important dates stay buried inside those files: a membership that renews automatically unless it is cancelled 30 days earlier, a 24-month warranty that quietly runs out, a lease with a notice period. People usually notice them after the date has passed.

What it does

Renewal Deadline Radar is an orchestration Skill for UGREEN AI NAS. The user names the folders (and, optionally, a receipt album) that the Skill may read. The Skill then:

  1. finds candidate documents in those folders only (file-agent),
  2. extracts every renewal, expiry, warranty and notice-period term with a verbatim quote and page number (document-agent),
  3. optionally reads photographed receipts and warranty cards with OCR (photo-agent),
  4. computes the act-by date: the last day the user can still act (for example renewal date minus the notice period, or purchase date plus warranty length),
  5. shows one radar grouped into Overdue, Act within 30 days, Act within the horizon, Later and Needs review, with a source for every line,
  6. after explicit approval, tags the urgent files and saves the radar as a Markdown note (file-agent, knowledge-agent).

On the next run it reads the previous radar and marks items New, Changed or Resolved.

How we built it

The Skill follows the Uliya Skill Development Standard from the official UGOS Skill Creation Tool (uliya-skill-kit 0.7.3): one SKILL.md entry point plus four Markdown references (date rules, report format, test plan, disclosures). Every capability was checked against the official capability catalog (version 2.0.0), and the parent workflow only delegates to registered experts. The package passes the tool's static check with 0 errors and 0 warnings.

Design choices

  • Read-only by default; the only writes (one tag, one new note) need user approval. Nothing is moved, renamed, deleted or shared.
  • Never guesses: ambiguous dates such as 03/04/2027 go to Needs review with both readings.
  • Privacy: account, policy and ID numbers are masked in the radar; the AI NAS local model is recommended so paperwork stays on the device.
  • All five delegation states (completed, failed, cancelled, waiting_user, waiting_approval) are handled.

Challenges

Round 1 has no NAS access, so the real expert outputs (scanned PDFs, receipt OCR quality, approval flows) still need validation. The test plan in the package defines two core cases and an edge case for the finalist test day.

AI tools disclosure

This project was built with an AI coding assistant: Claude (Anthropic), used through the Aside AI browser assistant. It read the rules and starter kit, proposed the concept, drafted the Skill files and the proposal, and ran the static checks. Jaeyoung Kim commissioned the project, set the requirements, and reviewed and approved the submission.

Built With

Share this project:

Updates

Submission history