# Progressive Disclosure Progressive disclosure exposes a small amount of high-value information first, then loads implementation detail only when it becomes relevant. In AI systems this conserves finite context and prevents unrelated instructions, references, or code from competing for attention. Shazam's façade and one-capability-per-file layout are structurally analogous: the public name is visible while the module is loaded on demand. XDO's references and targets allow details to remain addressable without being inlined everywhere. Related: [[wiki/Just-in-Time Context|Just-in-Time Context]], [[wiki/Capability Discovery|Capability Discovery]], [[wiki/Filesystem-Based Agent Skills|Filesystem-Based Agent Skills]]. Sources: [OpenAI — Agent Skills](https://developers.openai.com/codex/skills), [Anthropic — Agent Skills](https://docs.anthropic.com/en/docs/agents-and-tools/agent-skills/overview)