Skip to main content

SeedStacking™ Resources

Workshop: The Hidden Language of AI
Three hands-on activities to go from reading about file types to actually using them.

Today's Daily Seed breaks down the file types behind every AI tool: Markdown, JSON, YAML, CSV, and plain text. These are not code. They are structured text formats designed for humans to read. And understanding them is the fastest path from "I use AI" to "I build with AI."

Here are three activities you can do right now. No software to install. No coding required.

Activity 1: The Markdown Challenge (Seed level, 5 minutes)

Take this plain paragraph:

"Professor Dean teaches AI literacy at Harvest Kernel. His methodology called SeedStacking helps educators and professionals build AI skills one layer at a time. The community has over 300 members and offers free courses on tools like Google Gemini."

Rewrite it using Markdown formatting. Add at least: one heading (# followed by text), bold text (wrap in double asterisks), italic text (wrap in single asterisks), and a bullet list (dash followed by a space).

You can test your work in any free online Markdown editor. Search "Markdown live preview" and paste your version in.

Activity 2: The JSON Decoder (Sprout level, 8 minutes)

Read this JSON and answer four questions:

{
"recipe": {
"name": "Classic Tomato Soup",
"prep_time": "15 minutes",
"cook_time": "30 minutes",
"servings": 4,
"ingredients": ["6 tomatoes", "1 onion", "2 cloves garlic", "2 cups vegetable broth", "1 tbsp olive oil"],
"difficulty": "beginner"
}
}

Questions: (1) What is the recipe name? (2) How many servings? (3) How many ingredients are listed? (4) What is the difficulty level?

Then try modifying it: change the recipe to something you would actually cook. Edit the name, ingredients, and difficulty. Keep the curly braces, square brackets, and commas in place.

Activity 3: Build Your First Skill Outline (Grow level, 10 minutes)

This is the real unlock. Write a skill outline for an AI assistant using the template below. Think about something repetitive in your work: a report you write weekly, a process you explain to new hires, an email format you reuse constantly.

---
name: my-skill-name
description: >
[2-3 sentences: what does it do, who does it help?]

[Your Skill Name]

Purpose

[What problem does this solve?]

When to Use

Triggers when the user says:

  • "[trigger phrase 1]"
  • "[trigger phrase 2]"
  • "[trigger phrase 3]"

What It Produces

  • [Output 1]
  • [Output 2]
  • [Output 3]

That template is Markdown (the structure) with YAML (the header between the dashes). Two file types. Zero code. And that is genuinely how custom AI skills are built.

Share your work below. Post your Markdown formatting, your modified JSON recipe, or your skill outline idea. The best skill ideas get featured in the community spotlight.

  • 1 comment

Dean Le Blanc

Activity 4 just dropped for this workshop. This one is the Harvest level.

Activity 4: Build a Complete, Functional Skill File (Harvest level, 20 minutes)

Take the skill outline you wrote in Activity 3 and turn it into a complete, functional SKILL.md file. This means:

  1. Flesh out the Purpose section with a real paragraph explaining the problem it solves and who it helps.
  2. Expand the "When to Use" section with at least 5 trigger phrases someone might actually say.
  3. Add an "Input Methods" section describing what the user provides (a URL, a topic, pasted text, etc.).
  4. Write a "What It Produces" section with specific, detailed output descriptions.
  5. Add at least one "Edge Case" that explains what the skill should do when something unexpected happens.

The test: could another person read your skill file and understand exactly what it does, when it fires, and what it produces without asking you a single question? If yes, you just built a real AI skill. Post it in the comments. The strongest submissions get featured in the community spotlight.

Here is the progression you just completed:

Seed: You learned Markdown formatting.
Sprout: You read and modified JSON.
Grow: You wrote a skill outline using both formats.
Harvest: You built a deployable skill file and shared it with the community.

That is SeedStacking in action. One layer at a time.