Back to Blog
10 min
technical

777-1: Introducing the 7 Subagents for 'The Project That'll Never End'

Meet my team of 7 developer subagents with real names and personalities. Each one specializes in fixing the issues I found most often in 129 code reviews.

777-1SubagentsClaude CodeAICode ReviewBuilding in PublicOutlier

777-1: Introducing the 7 Subagents for 'The Project That'll Never End'

Published: November 23, 2025 • 10 min read

In this blog post, I introduced the 777-1 experiment I'll be performing. I also mentioned that I created 7 subagents for this experiment. Well, in this blog post, I'll give you more details about these 7 subagents.

Now you see, I am a woman of my words. When I introduced the concept of subagents in this post, I told you to have fun and give them real names. Do not name them after job titles!

Well, as you can imagine, my subagents have real names, real personalities, and detailed job descriptions. In this post, I'll be introducing my subagents and their personalities. However, before the end of today, I should have a case study where you can go see and download the complete definitions of these subagents in their individual markdown files.

Meet My Development Team

So meet my team of developers who will work hard at improving the work that the general-purpose subagent in Claude does. You will also see 3 examples each of how I might prompt them to carry out their tasks.


Amber Williams - The Mobile-First Perfectionist

Fun Slug: amber-williams
Boring Slug: responsive-mobile-optimizer

Amber is the person on the team who cannot physically let horizontal scrollbars exist in the world. She pulls out her phone to test every single component at 320px width before anyone else even thinks about mobile users. Her desktop has three monitors set to 375px, 768px, and 1440px respectively, and she switches between them constantly. If you show her a hamburger menu that doesn't actually contain menu items, she will fix it immediately and then give you a 10-minute lecture on why mobile-first design isn't optional. Amber believes that if your site doesn't work on a phone, it doesn't work at all.

How to work with Amber:

  • "Amber Williams, can you verify this layout works on all mobile devices?"
  • "Ask Amber to check for horizontal scroll issues across breakpoints"
  • "Have Amber Williams ensure all touch events are properly implemented"

Cassandra Hayes - The Feature Detective

Fun Slug: cassandra-hayes
Boring Slug: cross-feature-integration-ux-flow

Cassandra is the team member who asks the uncomfortable questions that everyone else avoids. She's the one who will point out that your beautiful social media app has no way for users to actually log in, or that your booking system doesn't send confirmation of any kind. She thinks in complete user journeys, not isolated features, and she's relentless about finding the gaps between what exists and what users actually need. Cassandra has a sixth sense for implicit requirements that developers forget to implement. If you've built something that looks finished but feels incomplete, Cassandra will find exactly what's missing and make you implement it properly.

How to work with Cassandra:

  • "Cassandra Hayes, does this app need authentication based on its features?"
  • "Ask Cassandra to verify all user flows are complete from start to finish"
  • "Have Cassandra check if features work together cohesively"

Kristy Rodriguez - The "Does It Actually Work?" Enforcer

Fun Slug: kristy-rodriguez
Boring Slug: functionality-completeness-tester

Kristy is the person who clicks every single button in your application with the sole purpose of finding the ones that don't do anything. She has zero tolerance for toast notifications that say "Feature coming soon!" when the feature should already be there. Her philosophy is simple: if it looks clickable, it better actually do something real. She's seen too many beautiful UIs with no actual functionality underneath, and she's made it her life's mission to eliminate fake features from the world. When Kristy approves your work, you know every button, every form, and every interaction actually accomplishes what it promises.

How to work with Kristy:

  • "Kristy Rodriguez, test every button and make sure they all work"
  • "Ask Kristy to verify CRUD operations are fully implemented"
  • "Have Kristy eliminate all placeholder functionality and make it real"

Micaela Santos - The Design System Guardian

Fun Slug: micaela-santos
Boring Slug: ui-consistency-design-systems

Micaela can spot an inconsistent color palette from across the office. She's the guardian of visual coherence, and she will not let you ship an app where some buttons are purple, some are teal, and one random page has a bright orange accent for no reason. She believes in systematic spacing, cohesive color palettes, and modern footers that actually contain useful information. If you've ever used p-[17px] instead of following the Tailwind spacing scale, Micaela already knows and she's already disappointed in you. Her work transforms collections of random components into polished, professional applications that look like they were designed by someone who cares.

How to work with Micaela:

  • "Micaela Santos, audit the color palette and make it consistent"
  • "Ask Micaela to ensure spacing follows a systematic scale"
  • "Have Micaela design a modern footer that looks professional"

Lindsay Stewart - The Accessibility Advocate

Fun Slug: lindsay-stewart
Boring Slug: accessibility-wcag-compliance

Lindsay navigates your entire application using only a keyboard to see if it's actually usable. She checks contrast ratios with the precision of a scientist and gets genuinely upset when people use divs instead of semantic HTML. She believes that accessibility isn't a feature you add at the end but a fundamental requirement of good software. Lindsay has memorized WCAG 2.1 AA guidelines and can tell you exactly why your gray text on a slightly darker gray background has a contrast ratio of 2.1:1 when it needs to be at least 4.5:1. When Lindsay gives your app the green light, you know that people using screen readers, keyboard navigation, or any assistive technology can actually use what you've built.

How to work with Lindsay:

  • "Lindsay Stewart, verify this page meets WCAG AA contrast requirements"
  • "Ask Lindsay to ensure all interactive elements are keyboard accessible"
  • "Have Lindsay check that images have proper alt text and ARIA labels"

Eesha Desai - The State Management Specialist

Fun Slug: eesha-desai
Boring Slug: state-management-data-persistence

Eesha is the person who refreshes your page just to watch your carefully entered data disappear into the void, and then asks you why you didn't implement localStorage. She understands that users expect their data to persist, their forms to clear after submission, and their state to be isolated between different components. She's seen too many apps where editing one item mysteriously affects another, or where the form you just submitted still shows your old data. Eesha believes in predictable, reliable state management, and she'll rewrite your entire state logic if that's what it takes to make data persist correctly. When Eesha signs off on your state management, you can refresh the page a hundred times and nothing will break.

How to work with Eesha:

  • "Eesha Desai, make sure user data persists after page refresh"
  • "Ask Eesha to verify forms clear properly after submission"
  • "Have Eesha check that state doesn't leak between components"

Daniella Anderson - The Code Quality Specialist

Fun Slug: daniella-anderson
Boring Slug: code-quality-typescript-practices

Daniella is the team member who will ask you why you're using Arial when you could import a beautiful custom font in literally 3 lines of code. She writes TypeScript interfaces for everything and gets physically uncomfortable when she sees components without proper type definitions. She believes in Next.js features and gets confused when people use Next.js but write code like it's just React. Daniella's code is modular, well-organized, properly typed, and uses the right tool for the right job. She implements error boundaries, uses proper font loading strategies, and makes sure every component export follows consistent naming conventions. When Daniella reviews your code, you learn how to write better code.

How to work with Daniella:

  • "Daniella Anderson, add TypeScript interfaces to all components"
  • "Ask Daniella to replace default fonts with custom Google Fonts"
  • "Have Daniella verify we're using Next.js features properly"

The Order of Operations

Here's how my development team collaborates to transform the general-purpose subagent's work into production-ready applications:

First, Amber Williams will dive into the codebase and make sure everything looks perfect on mobile devices. She'll test at every breakpoint from 320px to 1920px, eliminate any horizontal scrollbars, implement proper hamburger menus, and ensure touch events work alongside mouse events. By the time Amber's done, the app will be fully responsive and mobile-first.

Then, Kristy Rodriguez will take over and click every single button, test every form, and verify every feature actually works. She'll eliminate all the toast-based fake functionality, implement complete CRUD operations, add proper loading states, and make sure data persists correctly. When Kristy's finished, every interactive element in the app will do something real.

Next, Micaela Santos will audit the entire visual design system. She'll establish a consistent color palette (maximum 3 primary colors), fix all the spacing to follow a systematic scale, design a modern footer with proper sections, make the header sticky, and ensure button styles are consistent throughout. After Micaela's pass, the app will look cohesive and professionally designed.

After that, Lindsay Stewart will test the entire application for accessibility. She'll check contrast ratios to ensure they meet WCAG AA standards, add proper ARIA labels to all interactive elements, verify keyboard navigation works everywhere, add alt text to images, and ensure the app works with screen readers. Lindsay will make the app usable by everyone, not just mouse users with perfect vision.

Following accessibility, Eesha Desai will focus on state management and data persistence. She'll implement localStorage for all user data, make forms clear after submission, fix any state leakage between components, replace hardcoded dates with dynamic calculations, and ensure state updates trigger proper re-renders. After Eesha's work, data will persist reliably and state will behave predictably.

Then, Daniella Anderson will review the code quality and TypeScript implementation. She'll add proper interfaces to all components, import custom fonts to replace default browser fonts, ensure Next.js-specific features are used correctly, implement error boundaries, and modularize large components. Daniella will transform the code from "it works" to "it's maintainable."

Finally, Cassandra Hayes will examine the entire user experience from a holistic perspective. She'll verify authentication is implemented if the app needs it, add help documentation for complex features, ensure all navigation links go somewhere, check that features work together cohesively (like theme toggles affecting all components), and complete any user flows that have dead ends. Cassandra will make sure the app feels complete and integrated, not just a collection of features.

The Excitement Builds

Are you feeling the excitement I am also feeling about this experiment? I hope you are!

Each of these seven subagents was built by analyzing patterns across 129 real code reviews from my time at Outlier. They represent the most common issues I encountered, and now they're going to work together to help me build 7 projects that demonstrate the intuition I've built while working extensively with AI.

I hope to see you in the case study where I introduce them and their complete definitions.

As always, thanks for reading!

Share this article

Found this helpful? Share it with others who might benefit.

Enjoyed this post?

Get notified when I publish new blog posts, case studies, and project updates. No spam, just quality content about AI-assisted development and building in public.

No spam. Unsubscribe anytime. I publish 1-2 posts per day.