Explain Skills and Progressive Disclosure in an Agent

Describe the concept of a skill within an agent architecture, contrast it with a standard prompt, and discuss how progressive disclosure can structure the information provided to the agent.

Constraints & Assumptions

Assume a generic architecture: a skill is a reusable task bundle that includes instructions and may also contain references or tools. Do not rely on any particular product's file format or loading mechanism.

Clarifying Questions

How are skills discovered and chosen? What information is loaded at the start? Who is responsible for permissions and validation? What occurs if a referenced document is untrusted or out of date?

What a Strong Answer Covers

Reusable procedures compared to per-interaction instructions, staged loading, tradeoffs in context and selection, and the boundaries of what a skill can guarantee.

Follow-up Questions

How can an excess of skill descriptions impair selection? At what point should detailed references be loaded? Why does bundling a procedure not grant authority to every instruction in a retrieved document? Overview: Explain reusable agent skills, prompt context, progressive disclosure, selection metadata, referenced resources, and the limits of permissions and validation.