Inspiration
Technical project documents frequently contain conflicting evidence, incomplete validation records, and unsupported performance claims. AI summaries can hide these differences instead of making them visible.
What it does
Evidence Clarity is a proposed UGREEN AI NAS Skill that examines files within an explicitly authorized folder, identifies contradictions and missing verification, and generates a source-linked evidence report without silently modifying original documents.
How it is designed
The Skill uses a structured workflow for access authorization, file inventory, evidence extraction, claim comparison, uncertainty classification, and approval-gated report generation.
Each conclusion must remain traceable to its source. Missing evidence is reported as missing rather than interpreted as successful verification.
Challenges and lessons learned
A plausible AI-generated answer is not equivalent to independently verified evidence. The design prioritizes traceability, least-privilege access, and transparent handling of uncertainty.
Development status
The initial Skill specification, Round 1 proposal, and pre-test ZIP package have been prepared. Official Uliya runtime validation and testing on UGREEN AI NAS hardware have not yet been completed.
Personal contribution
I am an independent technology researcher working in embedded systems, real-time Linux, FPGA/RTL, evidence integrity, and reproducible performance verification. I define the technical requirements, design the architecture, direct AI-assisted implementation, and assess experimental results.
Evidence Clarity applies these research principles to trustworthy AI-assisted file analysis.
Accomplishments that we're proud of
What we learned
What's next for Evidence Clarity
Built With
- ai
- ai-assisted
- evidence
- markdown;
- nas;
- skill
- skill.md;
- standard;
- ugos;
- uliya
