# Hung-yi Lee Skill

An agent skill distilled from a specific professor's teaching style, built from hundreds of lecture transcripts, a knowledge graph of his course material, and an authorized interview used to calibrate and correct the skill's assumptions.

This skill distills the teaching style of a named professor, not a collection of quotes but an attempt at his knowledge and way of thinking, built from 478 YouTube videos, 27 full transcripts, eight topic pages, and four research references, organized into a 916-node knowledge graph with ten concept communities. The description of his teaching style is used with his authorization, and the skill imitates his teaching method rather than impersonating him personally.

The skill was built in three steps: mining the transcripts for frequency-based speech patterns using another skill called Fable 5; using Fable 5 to run a multi-agent workflow that generated an interview script of reaction and reversed-generation questions grounded in the source material; and actually interviewing the professor with that script, with his answers becoming the highest-authority layer in the skill's configuration, overriding anything inferred from the transcripts alone where the two disagree.

It installs by cloning the repository into a location an AI coding assistant reads, such as its skills directory, so the assistant can load the SKILL.md file.

Website: https://github.com/voidful/hung-yi-lee-skill