fire-learncoding

/fire-learncoding

Turn any codebase into a linear learning walkthrough. Anti-vibe-coding. Anti-cognitive-debt. Grounded in Simon Willison's Agentic Engineering Patterns (simonwillison.net/guides/agentic-engineering-patterns/)

---

The Philosophy

"If you don't understand the code, your only recourse is to ask AI to fix it for you — like paying off credit card debt with another credit card." — Simon Willison

This mode prevents cognitive debt by walking through every file from the entry point outward, explaining WHAT each piece does, WHY it's written that way, and WHICH pattern it uses. Real code is extracted via shell tools (grep, cat, sed) — never paraphrased from memory.

---

Arguments

arguments:
  action:
    required: false
    type: string
    options: [on, off, --from-github, --from-path]
    description: "Toggle mode or load source"

  --from-github:
    type: string
    description: "GitHub repo URL to learn from"
    example: "/fire-learncoding --from-github https://github.com/user/repo"

  --from-path:
    type: string
    description: "Local path to learn from"
    example: "/fire-learncoding --from-path ./src"

  --watch:
    type: boolean
    default: true
    description: "Mode 1: Agent explains + scaffolds. You read and say 'next'."

  --active:
    type: boolean
    default: false
    description: "Mode 2: Agent explains purpose, you write the key logic."

  --step:
    type: integer
    description: "Jump to a specific step (resume)"
    example: "/fire-learncoding --step 5"

  --entry:
    type: string
    description: "Override entry point detection"
    example: "/fire-learncoding --entry src/server.ts"

---

Mode Reference

Mode	Flag	You do	Agent does
Watch	--watch (default)	Read + say "next"	Explains, scaffolds file
Active	--active	Write key logic sections	Explains purpose, marks // WRITE THIS:
Hybrid	--hybrid (future)	Fill in business logic TODOs	Scaffolds all boilerplate