Mastery Module 9 · Let the LLM Set Things Up

Project Templates

Last reviewed

Beginner

What you'll learn

~15 min
  • Turn an idea into a short starter goal plus a check for when it is done
  • Use Plan mode so the agent asks questions and picks sensible tools
  • Choose the right starter card for your goal (portfolio, dashboard, tool, research)
  • Adapt a starter card for biotech, business, or government work

By the end of this lesson, you will have a set of short starter goals to adapt for your own projects. Each one pairs a goal with a check for when it is done. Paste one into your AI tool (such as Claude Code, Antigravity, or your preferred tool) in plan mode and let it ask you questions.

The Mental Model: Fill-in-the-Blank Recipes

A recipe card tells you what the dish is and how to tell it came out right. It does not tell you which pan to use. These cards work the same way: you state the goal and the check, and the agent chooses the tools after asking you what it needs.

The Starter Card

Every card has two parts:

Starter goal: [what it is, who uses it, the main things it does]
What to check when done: [one or two things you can click or see]

Switch your AI tool to plan mode inside the session, send the goal, answer the agent’s questions (look and feel, how much content, any tools you care about), and approve the plan. Then use the check to decide whether the result is right.

💡Not sure about tools?

Leave them out. The agent will propose sensible choices in its plan and ask if you have a preference. Name a tool only when you already know you need it.

ℹQuick glossary

Hero = the large banner at the top of a page. CTA = “call to action,” a button that asks the visitor to do something. KPI = key performance indicator, a number that shows how something is going. WCAG = Web Content Accessibility Guidelines, the standard for sites people with disabilities can use.

Personal / Portfolio

Personal Portfolio Website

Starter goal: A personal portfolio site with a hero, an about section, three featured projects, and contact info. Dark and clean, with clearly labeled placeholder text I can replace.
What to check when done: It looks right on a phone-sized window, and each project card opens its own page.

Personal Blog

Starter goal: A blog with a home page of recent posts, a page for each post, tags that filter posts, and an about page. Easy to read, with three sample posts.
What to check when done: Clicking a tag shows only posts with that tag.

Business / Professional

Landing Page

Starter goal: A product landing page with a headline, two buttons, six features, three testimonials, three pricing tiers, and a short FAQ. Placeholder text that looks realistic.
What to check when done: Each FAQ question opens and closes, and the page reads well on a phone.

Dashboard

Starter goal: An admin dashboard with a side menu, an overview page of stat cards and a chart, a searchable users table, and a settings form. Use realistic mock data.
What to check when done: Sorting and searching the users table changes the rows, and the side menu collapses on a phone.

Tools / Utilities

Todo / Task Manager

Starter goal: A task manager where I can add, finish, edit, and delete tasks, group them by project, and filter by done or not done. Tasks should still be there after a refresh.
What to check when done: Add three tasks, finish one, refresh the page, and confirm all three are still there.

Calculator / Converter

Starter goal: A unit converter for length, weight, and temperature that updates as I type and lets me swap the direction.
What to check when done: 100 Celsius shows 212 Fahrenheit, and swapping the direction gives 100 back.

Research / Academic

Data Visualization Page

Starter goal: A page where I upload a CSV and see basic statistics and a few charts, with sample data built in.
What to check when done: The sample data loads, and my own small CSV shows the right row count.

Note-Taking App

Starter goal: A note-taking app with folders, search, and Markdown formatting. Notes should be saved on my computer and still be there tomorrow.
What to check when done: Write a note with a heading and a list, search for a word in it, and confirm it appears after a refresh.

Python Data Pipeline

Starter goal: A Python tool that reads a CSV, cleans missing values, summarizes each column, and saves a chart as an image. Include a small sample file.
What to check when done: Running it on the sample file produces a summary and an image file.
💡How to customize these cards

Add who uses it, the real column or field names, and anything that must be true (for example, accessibility). Cut anything you do not care about. If you are unsure what to add, the agent’s questions will draw it out.

Worked Example: Adapting a Card for Your Domain

Take the generic starter card for a data page:

Starter goal: A page where I upload a CSV and see basic statistics and charts.
What to check when done: My own small CSV shows the right row count.

Biotech:

Starter goal: A gene expression explorer. I upload a CSV of gene IDs, fold change, and p-values, then see a volcano plot and filter by significance. Include sample data.
What to check when done: Moving the p-value filter changes how many genes are marked significant.

MIS analyst:

Starter goal: A sales dashboard. I upload a CSV of date, region, product, and revenue, then see totals, revenue over time, and a regional comparison. Include 12 months of sample data.
What to check when done: The regional totals add up to the overall total.

Government analyst:

Starter goal: A public spending dashboard. I upload a CSV of department, category, amount, and fiscal year, then see totals and a searchable table. It must meet WCAG 2.2 AA, and I will check it with accessibility tools afterward.
What to check when done: I can reach every control with only the keyboard, and text contrast is high.

Same card, different details. The agent still asks about the rest.

🧬In Your Field: Biotechclick to expand

Biotech starter goals:

Plate layout planner:

Starter goal: A 96-well plate planner where I label wells with sample names and export the layout as a CSV.
What to check when done: Labeling A1 to A3 and exporting gives three matching rows.

Primer tool:

Starter goal: A primer calculator that takes a sequence and shows length, GC percent, and reverse complement, and remembers my last few pairs.
What to check when done: A known primer gives the GC percent I expect from a hand count.
📊In Your Field: MIS / Businessclick to expand

MIS / Business starter goals:

Client tracker:

Starter goal: A contact manager with card and table views, search, status tags (Active, Follow-up Needed, Inactive), and sorting by last contact date.
What to check when done: Sorting by last contact puts the stalest client first.

Project budget tracker:

Starter goal: A budget tracker for line items with estimated and actual cost, showing variance and red, amber, or green status, plus an estimated versus actual chart and CSV export.
What to check when done: An item 20 percent over budget shows red, and the totals match a hand sum.

Run this on your machine

Think of a small tool that would help your work or studies, such as a team project tracker.

  1. Open a fresh folder in your agent (Desktop App or CLI) and switch to plan mode.
  2. Write a starter goal and a check for it, in two or three sentences.
  3. Answer the agent’s questions about pages, features, and look.
  4. Approve the plan, let the agent build and preview the project, and run your check in the browser.
  5. If something is off, say so in plain words and let it fix it.

When Things Go Wrong

🔧

When Things Go Wrong

Use the Symptom → Evidence → Request pattern: describe what you see, paste the error, then ask for a fix.

Symptom
The AI built something completely different from what you described
Evidence
I asked for a dashboard and got a static landing page
What to ask the AI
"This isn't what I asked for. I need an interactive dashboard with a side menu, charts, and filters, not a static landing page. Show me a new plan first, and ask me anything you need."
Symptom
The project is missing features you specified
Evidence
I asked for search and filter functionality but the page just shows a static list
What to ask the AI
"The search and filters are missing. Please add a search bar that filters the list as I type, and dropdown filters for category and status."
Symptom
The generated code has errors and won't run
Evidence
The preview shows error messages or a blank page
What to ask the AI
"The preview shows errors. Here is the message: [paste error]. Please fix them so the project starts cleanly and every page shows up."
Symptom
The design looks nothing like what you asked for
Evidence
I asked for dark theme but got a white background with default styles
What to ask the AI
"The theme is wrong. I wanted a dark background with light text throughout. Please update the global styles and every component."

Key Takeaways

  • A starter card is a goal plus a check: what it is and how you will know it works
  • Use plan mode so the agent asks questions and chooses tools, instead of you pre-specifying a stack
  • These cards are starting points. Swap in your own field, data, and rules
  • Relevant detail helps. Real field names and must-have requirements beat decoration. Avoid conflicting or unnecessary constraints
  • When the output does not match the goal, refine and re-ask rather than starting from scratch
🔍Keeping the templates you reuseclick to expand

As you use these cards, you will develop favorites that work well for your field. When you reach for one again, keep it where your tool can run it, as a skill, a saved prompt file the agent runs by name (Giving Good Instructions’s rule: a reused prompt becomes a skill). A starter goal runs before the project exists, so it fits the personal skills folder that lesson points to.

ℹWhat to do when the generated app doesn't work

Scaffolded apps always need checking. If the app does not start or shows errors, paste the full error message back to your AI tool with “fix this.” If it used a library you did not want, say so: “Replace [unwanted library] with [preferred alternative].”

KNOWLEDGE CHECK

Which part of a starter card matters most for a good result?

KNOWLEDGE CHECK

You used a dashboard card but your field needs a 'compliance status' column in every table. What should you do?

💡What's next

Now that you can start projects from short goals, Module 10 teaches you how to plan multi-step builds and break large projects into manageable chunks.

Search lessons