Unverified60% confidenceFactExact time
Skills consist of two parts—Description and Content; the Description is unconditionally loaded into the context at the start of every conversation, while the Content is only expanded and read when the model decides to use that Skill.
2
Sources
60%
Confidence
Long-term
Relevance
8/25/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSkill internalization refers to the Executor gradually absorbing the behavioral patterns carried by external textual skills into its own policy during training, so that at inference time it can complete the same type of task even without skills being provided.80% similarVerifiedThe Skill identification process typically relies on the model's semantic understanding of capability descriptions, and how a Skill description is written has a greater impact on actual invocation outcomes than the complexity of the code implementation.79% similarUnverifiedSkills use a lazy-loading mechanism: at agent startup, only each Skill's name, one-sentence description, and local path are written into the context, and only after the model calls a file-reading tool such as Read does the Skill body get added to the context as tool return content.79% similarVerifiedThe essence of the Skill mechanism is to make implicit prompt knowledge explicit and versioned, so that it can be stored, shared, and iterated upon just like code.79% similarUnverifiedSkill.md differs from fine-tuning (which changes model weights) and prompt engineering (which relies on the context of a single conversation): it is a reusable asset independent of the model that can be loaded by an agent at runtime and followed step by step.78% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/797941API
curl https://kongchang.com/api/v1/knowledge/claims/797941MCP
get_claim(id=797941)