Video Knowledge Distiller

Problem It Solves

After reading a methodology-heavy book, video, or podcast, useful knowledge often stays descriptive: “I got the idea,” but not “when should I apply it, and how?” This skill turns high-value content into executable SKILL.md files for agents. Its goal is not to produce summaries, reflections, or author-style roleplay, but to extract reusable frameworks, principles, cases, counterexamples, and glossary terms.

How It Works

The skill follows a RIA-TV++ pipeline. It first performs Adler-style analysis of structure, terminology, argumentation, and application context. Then it runs parallel extractors for frameworks, principles, cases, counterexamples, and glossary items. Each candidate passes triple verification: independent evidence across contexts, predictive power, and non-obvious value. Accepted items become skills with six sections: original quote, interpretation, past application, future trigger, execution steps, and boundaries. The output also includes related_skills links, an INDEX.md graph, and pressure tests for trigger, non-trigger, and edge cases.

Boundaries And Notes

The skill requires supplied text, files, or video links. For video sources it prefers existing transcription, subtitles, or platform subtitles when available, and avoids guessing content. Dense methodological sources usually yield more usable skills than essays, interviews, or promotional material. If many candidates are rejected, that may indicate fewer reusable frameworks in the source or a need to adjust the verification threshold.