In this episode of AI That Works, we dive into the often confusing world of coding agents! We're breaking down the differences between commands, skills, agents, and sub-agents – all the pieces that fit into a coding agent harness like Claude Code.
We'll share our take on how to structure your agents for optimal performance, emphasizing the importance of removing irrelevant context ASAP. We'll also discuss why separating instruction modules from context isolation is crucial and why you shouldn't put custom instructions directly into your agents.
Plus, we touch on the dangers of over-relying on agent swarms and "lights off" software factories.
Check out our github: https://www.github.com/boundaryml/baml
AI That Works repo: https://github.com/ai-that-works/ai-that-works
Socials:
X: https://x.com/boundaryml
Discord: https://discord.com/invite/yzaTpQ3tdT
LinkedIn: https://www.linkedin.com/company/boundaryml/
⏱️ Chapters:
Intro
History of Subagents
Tool Calls and Token Usage
Launching Skills
Skills and Bundled Files
Internal Skills: Application Reliability
Skills vs MCPs
Sharing and Context Windows
Twitter and Cloud Code
Transcript
I've watched this one closely. Ask me anything about it and I'll point you to the exact moment.