Inspiration
We build AI Examiner, a study app used by over 20,000 students in Nigeria. Every exam season we see the same thing: course material is scattered across lecture PDFs, phone photos of the board and voice recordings, and students lose hours turning it into something they can revise from. Learners with dyslexia have it worse, because most material is dense text. We also kept hearing that students don't want to upload private course files to a cloud chatbot.
What it does
Study Pack is a UGREEN AI NAS Skill. A student names one folder (for example "Pharmacology / Week 6") and the Skill:
- lists the documents, photos and recordings in that folder only
- summarises each document with page references
- reads text from photos of notes or the board
- transcribes lecture recordings
- writes practice questions with an answer key that cites the source file and page
- writes an Easy Read version for learners who struggle with dense text
- saves everything as one Markdown study pack next to the material, after approval
Original files are only read, never changed.
How we built it
We used the UGOS Skill Creation Tool documentation and the public capability catalog to design a parent Skill that delegates to four registered experts: file-agent (ug-filesystem), document-agent (ug-document qa and writing), photo-agent (ug-photo vision-ocr) and knowledge-agent (ug-knowledge audio-transcription and save-text). Each step only receives real results from the step before, and every write goes through platform approval.
Challenges
Making the workflow safe and honest: no searching outside the folder the user named, stopping cleanly on an empty folder, saving partial packs when some files fail, and never inventing a summary or page reference that was not in the material.
What we learned
Delegation-based Skills force you to think about failure states first. Round 1 is a design and SKILL.md package; if we reach the finals we will test question generation and handwriting OCR on a real NAS.
What's next
Testing on a real UGREEN AI NAS, then adding support for more languages for students across Africa.
Built With
- artificial-intelligence
- markdown
- ocr
- speech-to-text
- ugos
- ugreen
