NexusGPT
Extending AI agent capabilities and
behavior through modular skill flow design
Overview
1/4
Problem
3/7
Nexus GPT simplifies agent customization by letting users build reusable skills that enhance task precision, speed, and autonomy
Nexus GPT focuses on making agents more adaptable and user-aligned. While agents could perform general tasks, they often lacked the domain-specific context users needed. The Skill Builder closes this gap by allowing users to define agent capabilities through simple, no-code configurations—improving agent output and reducing repeated prompt engineering.
Timeline
3 months (Jan '25 - Apr '25)
My role
Product designer
Team
1 PD, 1 PM, 2 DEVs
Status
Under-development
Collaborated closely with product and engineering to understand agent architecture and user expectations.
Designed the Skill Builder from the ground up using fast, iterative prototypes.
Led the design vision and execution as the sole designer on the project.
Background
2/4
Problem
3/7
Prompt-driven agents hit a ceiling when operating in isolation, which is why integrating skills is essential to get the most out of your agents
To truly replicate the capabilities of an employee, agents must interface with external tools, APIs, and systems. Nexus' Skill Builder unlocks this—enabling agents to not just reason, but to act meaningfully in the real world.
Problem
3/7
Ideal
Write a prompt. Done. That's the promise.
Reality
Users struggle to prompt. Agents underperform.
Problem
3/7
For example, writing an email is one thing—sending it via Gmail or referencing live data through Google Search is what makes it useful.
Problem
3/7
Current skill builder was underpowered and unintuitive
The existing Skill Builder was outdated in both functionality and clarity. Users struggled to parse the interface, understand the flow, or grasp what actions they were performing. It also lacked key features that could have significantly enhanced its utility and empowered users to build more capable agents.
Problem
3/7
Problem
3/7
Core challenge
How can we enable users to easily and effectively make their agents more capable and powerful?
Designs
2/4
Problem
3/7
Problem
3/7
Problem
3/7
Problem
3/7
Problem
3/7
