OC-201b · Module 3

Skill Composition

3 min read

Individual skills are useful. Composed skills are powerful. Skill composition is the practice of building complex automations by chaining simpler skills together, where each skill's output feeds the next skill's input. The content pipeline from OC-201a is a composition: the link-ingestion skill feeds the knowledge-base skill, which feeds the video-ideation skill, which feeds the task-creation skill. Four simple skills, chained, produce a workflow that turns a dropped URL into a ready-to-produce task card in 30 seconds.

Composition works because of contracts. Each skill promises a specific output format. The next skill expects a specific input format. When those formats align, the skills snap together like plumbing. When they do not align, you need an adapter — a thin transformation layer between two skills that converts one format to another. Adapters are small, testable, and replaceable. They are the joint compound between two pipes. The pipes do the real work. The adapters make the connections clean.

Do This

  • Build each skill to do one thing well with a clean input/output contract
  • Use adapters between skills whose contracts do not perfectly align
  • Test each skill independently before testing the composition
  • Document the composition as a chain: skill A → adapter → skill B → skill C

Avoid This

  • Build monolithic skills that handle the entire workflow in one function
  • Assume two skills will work together because they "probably" return similar formats
  • Test only the full composition — when it breaks, you will not know which skill failed
  • Embed transformation logic inside a skill to make it fit one specific composition