Case Studies

These cases show how the readiness gate behaves across source types. They are not claims that the generated artifacts are publication-ready without review.

Bundled Demo Article

source-to-skill analyze examples/article.md

The bundled article is a compact, structured source. It is useful for checking the basic flow, but it is intentionally small; it should not be treated as proof that every short source deserves a standalone skill.

Bundled Meeting Transcript

source-to-skill clean-transcript examples/meeting-transcript.vtt --out out/clean-meeting.md
source-to-skill split-source out/clean-meeting.md --out out/topics

The transcript example demonstrates the conservative path: clean noisy speech, split it into topic candidates, then build or fold only the useful segment. Ordinary meetings often make better notes or seeds than full skills.

Private Long-Form EPUB Validation

A privately supplied Chinese EPUB edition of The Psychology of Money was used as a local validation case. The book text is not included in this repository. Use your own legally obtained files for local analysis.

Observed readiness report:

Metric	Result
Score	88/100
Recommended output	Full Skill
Effective words	99,321
Headings	46
Source depth	25/25
Structure	20/20
Actionability	20/20
Transferability	8/15
Evidence	6/10
Signal cleanliness	9/10

This case supports the product principle: a substantial, structured book can clear the Full Skill threshold, while short meetings and noisy transcripts should usually stop at Note or Skill Seed.

Boundary

source-to-skill is currently strongest as a readiness gate and artifact scaffold. It can identify when a source looks skill-worthy, normalize supported input formats, and create reviewable files. It does not replace human review, copyright judgment, or source-specific editorial work.