Progressive disclosure means showing the model detail only when it needs it, instead of loading everything into context up front. You give the agent a short overview now and a way to fetch depth on demand.
Why not just include everything
Because context is not free. Every token you load competes for the same context window and spreads the model's attention thinner across everything else in there. A prompt stuffed with material the agent never touches on this task is pure cost: slower, pricier, and easier to get lost in.
In practice progressive disclosure looks like:
- A skill the agent loads only when the task calls for it.
- A context pointer (a path or link) it opens if it turns out to be relevant.
- A one-line summary now, with the full document a single tool call away.
The payoff
The agent works from a lean, relevant context and reaches for detail on its own terms. You get the reach of a large knowledge base without paying to keep all of it resident, and without burying the few facts that matter under the many that do not.
Related terms
Skill
A skill is a packaged, reusable set of instructions an agent loads on demand for a specific kind of task. It is progressive disclosure of know-how instead of cramming everything into the system prompt.
Read definition →Context pointer
A context pointer is a reference (a path, URL, or id) you give an agent instead of the full content, so it can fetch the material only if and when it needs it. It is a cheap way to make a lot of context available.
Read definition →Context window
The context window is the maximum amount of text, measured in tokens, that a model can consider for a single request. It is a hard ceiling, and it is the main resource you manage when working with an agent.
Read definition →