Unverified50% confidenceFactExact time
The second step is the host injecting the tool description text into the model's context; the model's entire knowledge of the tool comes only from this description.
1
Sources
50%
Confidence
Long-term
Relevance
10/21/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedHarness writes everything the model sees—including system prompts, chain-of-thought, tool calls, and sub-agent scheduling—into an append-only log.79% similarUnverifiedThe tool definitions themselves also consume context tokens, and the model's understanding of a tool depends on the clarity of its description.76% similarUnverifiedSeparating tool definitions from models means that the same set of tools can serve different models.72% similarUnverifiedLarge models determine what a skill or tool is for and when to call it by reading its description.71% similarVerifiedThe key breakthrough in tool-calling capability lies in the model's ability to generate structured function-call requests and correctly parse tool return results before continuing to reason, rather than simply concatenating tool outputs into the prompt71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1081505API
curl https://kongchang.com/api/v1/knowledge/claims/1081505MCP
get_claim(id=1081505)