I've been using the Gemini CLI for a while, mostly calling simple skills to handle routine workflow automation. Recently, I decided to push it further into actual software development. I came across a CLI extension called `conductor`—a tool that takes a written specification and magically generates code. Around the same time, I discovered dedicated `gemini` and `firebase` CLI skills designed to facilitate coding tasks.
This sparked a question: **Can the `conductor` extension and these agent skills actually complement each other to develop a full-stack application?**
In this session, I will share the unfiltered journey and raw feelings from that experiment. Instead of a perfectly polished demo, we will look at the reality of spec-driven, terminal-based AI development.
I’ll walk through how I used this combination of tools to build a simple but complete architecture:
1. Writing specs for `conductor` to trigger the `firebase` and `gemini` skills.
2. Generating a Python-based Firebase Cloud Function that calls the Gemini API.
3. Scaffolding an Angular application to serve as the user interface and display the generated responses.
