Inspiration
In a family emergency, important information may already exist in insurance policies, medical reports, contracts, and phone photos—but finding it quickly is difficult. Family Emergency Binder explores how a NAS can help families prepare before that moment arrives.
What it does
The Skill is designed to turn existing NAS documents and card photos into a cited, privacy-masked emergency binder. It includes an emergency card, a readiness grid for each family member, and flags for missing, expired, soon-expiring, contradictory, or unreadable records.
Three workflows are defined:
- Build: prepare a binder from user-confirmed sources.
- Refresh: compare against a previous binder and check source fingerprints.
- Ask: retrieve short answers to urgent questions with source citations.
A photo gap search ranks possible unfiled card images and explains the evidence behind each match.
How it was built
The deliverable contains SKILL.md and six reference documents. The workflow delegates file discovery, photo reading, document extraction, and saving to UGREEN's registered experts.
The instructions request NAS built-in models, mask sensitive identifiers, and require approval before saving or performing optional tagging and album operations.
OpenAI Codex assisted with drafting, documentation review, failure handling, and local static and packaging checks.
Challenges and lessons learned
Key challenges included handling conflicting sources, distinguishing missing information from negative findings, and designing fallbacks for uncertain runtime behavior.
The design emphasizes explicit evidence, bounded operations, and honest reporting when a feature is only partially available.
Current status
The local UGREEN kit checker reported static-pass with zero errors or warnings. ZIP integrity and agreement with the source files have been verified.
Actual NAS execution, extraction accuracy, photo recall, and runtime fallback behavior remain unverified.
What's next
Validate Build, Refresh, and Ask on the NAS using fictional document and card data, consenting participants for face-index photos, and cases covering expiry dates, conflicting records, unreadable files, and declined saves.
Built With
- markdown
- openai-codex
- ugreen-ai-nas
