Descriptions:
Nick Nisi and Zack Proser, developer experience engineers on WorkOS’s applied AI team, lead this hands-on workshop on building “skills” — reusable, structured context units for AI coding agents like Claude Code. The session addresses a fundamental limitation of agentic AI development: every conversation starts from zero, forcing developers to re-explain project conventions, tooling preferences, and coding standards each time. Skills offer a way to encode that institutional knowledge once and deploy it consistently across sessions, terminals, and team members — carrying the DRY principle into the agentic era.
The workshop covers how skills differ from static memory files like CLAUDE.md, when to invoke a skill via slash command versus letting the model pick it up automatically, and how to test and track skill reliability across model upgrades. Nisi and Proser draw on direct experience — both describe having largely stopped writing code manually over the past six to eight months — and walk through real WorkOS examples. They also address tricky edge cases raised by attendees: handling conflicting skills, reusing skills inside sub-agents, and deciding when a skill is the right abstraction versus building a CLI or spawning a sub-agent with its own context window.
The session includes a companion GitHub repository with slides and skill examples, making it straightforward for attendees to follow along or reproduce the patterns in their own codebases.
📺 Source: AI Engineer · Published May 06, 2026
🏷️ Format: Hands On Build
