I was using the term as you used in your comment. I believe the official term is "Activation" however:

> Activation: When a task matches a skill’s description, the agent reads the full SKILL.md instructions into context.[1]

> Full instructions load only when a task calls for them, so agents can keep many skills on hand with only a small context footprint.

[1]: https://agentskills.io/home#how-do-agent-skills-work

Concretely, it has to decide whether it is in a circumstance where that skill is useful, pull the instructions into the context and follow them.