Skill Creator Primer

Note: If your environment does not have the skill-creator skill: Stop and ask the user to run /plugin marketplace add anthropics/skills then skill-creator@claude-plugins-official before proceeding.

How Skills Actually Work

Understanding these mechanics helps you design more effective skills.

Skills are prompt-based context modifiers, not executable code. When invoked, a skill:

1. Injects instructions into the conversation context (via hidden messages to the agent)
2. Modifies execution context by changing tool permissions and optionally switching models
3. Guides the agent's behaviour through detailed instructions

Skill selection happens through pure LLM reasoning. No algorithmic matching, keyword search, or intent classification. The agent reads skill descriptions in the Skill tool's prompt and uses language model reasoning to decide which skill matches. This makes the description field the single most critical element.

Agents tend to under-trigger skills. To combat this, make descriptions slightly assertive about when to activate. Instead of "Build dashboards for data", write "Build dashboards for data. Use this skill whenever the user mentions dashboards, data visualisation, metrics, or wants to display any kind of data."

Progressive disclosure keeps context lean. Three-level loading: