Tutorial-to-Template — First Attempt

Convert a YouTube tutorial or transcript into an executable project template and an Obsidian-ready knowledge artifact. This repository contains an MVP CLI, extraction rules, and a regression fixture.

Example source video: https://www.youtube.com/watch?v=96jN2OCOfLs

Video title: Andrej Karpathy: From Vibe Coding to Agentic Engineering Channel: Sequoia Capital Speakers: Andrej Karpathy, Stephanie Zhan Duration: 00:29:49

What this run proves

This source is not a step-by-step coding tutorial. It is a methodology talk. The correct transformation is therefore not to invent a repository from thin air, but to produce an execution template:

• Obsidian note for durable understanding
• Agent context pack for implementation work
• Prompt/spec templates for agentic engineering
• Verification checklist to prevent vibe-only output
• Task list for turning the idea into a product MVP

Output files

tutorial-to-template-first-attempt/
  package.json
  src/
    cli.ts
    extract/
    render/
    schemas/
    io/
  fixtures/
    karpathy-agentic-engineering.txt
  tests/
    karpathy-fixture.test.mjs
  output/
    karpathy-agentic-engineering/
      README.md
      PROJECT_TEMPLATE.md
      TASKS.md
      OBSIDIAN_NOTE.md
      AGENT_CONTEXT.md
      TEMPLATE.json
  README.md
  generated-template/
    PROJECT_TEMPLATE.md
    TASKS.md
    TEMPLATE.yaml
  obsidian/
    Andrej Karpathy - From Vibe Coding to Agentic Engineering.md
  agent-context/
    AGENT_CONTEXT.md
    CODING_AGENT_BRIEF.md
  prompts/
    extraction-prompt.md
    agent-review-prompt.md
  schemas/
    tutorial-template.schema.json

Product principle

A saved tutorial is passive. A generated template is executable.

For non-code videos, the system should emit a concept_to_execution_template rather than hallucinating code.

CLI MVP

Run the first regression fixture:

npm run generate:karpathy

Or run the CLI directly with a transcript file:

node src/cli.ts \
  --url "https://www.youtube.com/watch?v=96jN2OCOfLs" \
  --transcript-file fixtures/karpathy-agentic-engineering.txt \
  --out output/karpathy-agentic-engineering

The implementation follows:

input -> classify -> extract -> validate -> render -> write_to_vault

For an Obsidian vault copy, add --vault /path/to/vault/folder.

Regression test

npm test

The test asserts that the Karpathy fixture is classified as concept_talk, install commands remain not_specified, recommended defaults are separated, and all expected Markdown/JSON artifacts are written.