777-1: GoalStack (7 Subagents, 7 Predictions, 0 Faith)
Published: December 10, 2025 - 5 min read
Welcome to Project 2 of the 777-1 experiment! If you missed Kinetic Canvas, go check that out first. Otherwise, let's keep this train moving.
Today's project has something special: authentication. This is the only project in the entire series where I explicitly mention login functionality in the prompt. Guest mode, logged-in mode, data that needs to persist across sessions. My subagents have... thoughts.
What is GoalStack?
GoalStack is a certification and goal tracking tool where users break down large objectives into actionable steps and visualize their progress through charts and motivational feedback.
Think of it like this: you have a big goal (get AWS certified, learn Spanish, finally understand Kubernetes). You break it into milestones. You check off milestones. You watch a pretty progress bar fill up. The app celebrates your wins with encouraging feedback because we all need a little dopamine hit when we're grinding through certification prep.
Application Category: Productivity
Complexity Tier: Medium
The Prompt: Finding the Goldilocks Zone
The starting prompt I'll be using for this application is:
I want to create a certification and skills tracker called GoalStack. The idea is that people can add big goals they're working toward—like getting a new certification, learning a language, or developing a professional skill—and then break each goal into smaller milestones they can check off along the way. Show their progress visually with charts or progress indicators. When users complete milestones, give them some kind of encouraging feedback. The app should load with sample data so first-time visitors can see how it works without signing up, but they'd need to log in to create and track their own goals. Skip the registration flow since this is a demo.
Use a neumorphism design style with #7C3AED (violet) as the primary color, #E9D5FF (lavender) as secondary, and #F472B6 (pink) as an accent.
Again, this is my attempt at finding the goldilocks zone for context engineering. Not too sparse, not too overwhelming. The big difference here? I'm explicitly asking for guest mode AND login functionality. The transition between these states is where things get interesting (and where Cassandra Hayes starts sharpening her knives).
Meet the Critics
You already know my seven subagents from Project 1, so I won't repeat the full introductions. Quick refresher:
- Amber Williams - Mobile-First Perfectionist
- Kristy Rodriguez - "Does It Actually Work?" Enforcer
- Micaela Santos - Design System Guardian
- Lindsay Stewart - Accessibility Advocate
- Eesha Desai - State Management Specialist
- Daniella Anderson - Code Quality Specialist
- Cassandra Hayes - Feature Detective
The Predictions
I gave them the prompt. I asked them to imagine what the general-purpose subagent would build. Here's what each one expects to find:
Amber Williams (Mobile-First Perfectionist):
The goal cards look beautiful at 1440px. At 375px, they're playing a game of Tetris that everyone loses.
Kristy Rodriguez (Functionality Enforcer):
There's a 'Try as Guest' button that leads to a page with the text 'Guest mode coming soon.' I admire the optimism.
Micaela Santos (Design System Guardian):
Every progress bar has chosen its own shade of green. They're all wrong, and none of them match each other.
Lindsay Stewart (Accessibility Advocate):
The progress percentage updates visually but never announces to screen readers. Blind users will simply have to guess how close they are to their goals.
Eesha Desai (State Management Specialist):
Complete a step. Watch the parent goal ignore this information entirely. Refresh. Watch your steps forget they were ever completed.
Daniella Anderson (Code Quality Specialist):
Goals contain steps. Steps contain substeps. None of these have TypeScript interfaces. The nesting goes five levels deep. I'm going to need therapy.
Cassandra Hayes (Feature Detective):
You can spend an hour in guest mode adding goals. Then you sign up. Your goals? Gone. The app has taught you an important lesson about data migration.
Zero Faith, Maximum Entertainment
Cassandra's prediction is BRUTAL. And honestly? It highlights exactly what makes this project interesting from a prompt engineering perspective. I told the AI about guest mode and login mode, but I never explicitly said "migrate guest data when they sign up." That's an implicit requirement. A human developer would probably ask about it. Will the general-purpose subagent figure it out?
Eesha and Daniella are also circling the same concern from different angles. Hierarchical data (goals containing steps containing substeps) is notoriously tricky to manage. State updates need to cascade properly. TypeScript interfaces need to capture the nesting. And persistence needs to work across both guest and authenticated states.
This is the kind of complexity that looks simple on the surface. "Just track some goals," you might think. And then you realize you've built a nested state management nightmare with authentication-dependent storage and you're crying into your keyboard at 2 AM.
What's Next
Stay tuned for the next post where I introduce MotorMatch, a peer-to-peer vehicle marketplace. My subagents are already concerned about the lack of authentication in that one. Apparently there's no middle ground.
As always, thanks for reading!