Skills at Scale — Nick Nisi and Zack Proser, WorkOS

Skills at Scale focuses on enhancing AI agent performance by creating modular, portable "skills"—discrete units of work that combine markdown instructions with deterministic script execution. By moving beyond static memory files like `Claude.md`, developers can inject hyper-specific context and logic only when needed, reducing context window bloat and improving reliability. Key strategies include using progressive disclosure to load relevant documentation on demand, implementing confidence scoring to force iterative clarification, and leveraging script interpolation to ensure consistent data retrieval. These skills allow teams to codify workflows—such as repository health analysis, automated documentation, or complex API orchestration—into shareable, version-controlled units. This approach replaces generic, non-deterministic LLM interactions with structured, repeatable processes that maintain high accuracy across diverse codebases and team environments.