Skills

Skills are the building blocks of agentic workflows. A skill is a self-contained instruction set that tells an AI agent what to do, how to do it, and where to find supporting data.

Think of a skill as a runbook for an agent — structured enough to be deterministic, flexible enough to handle edge cases.

Why Skills Matter

Without skills, agents rely entirely on their training data. That means inconsistent outputs, hallucinated steps, and no access to your internal tools or APIs. Skills solve this by giving agents:

• Explicit instructions — step-by-step procedures instead of guessing
• Access to external resources — API scripts, reference docs, calculation engines
• Consistent behavior — same input produces same output, every time
• Progressive loading — only the relevant knowledge gets loaded, keeping context windows lean

Skill Architecture

Click any element to learn more about it.

Guides

• Anatomy of a Skill — The three-tier structure: metadata, instructions, supporting files
• YAML Frontmatter — How agents discover and trigger skills
• Writing Steps — Structuring the instruction body for reliable execution
• References — Storing large docs, examples, and static data
• Scripts — Running code for calculations, API calls, and file processing
• Progressive Loading — How the three tiers keep context windows efficient
• Testing & Iteration — Eval-driven development, benchmarking, and description optimization