Unverified50% confidenceCautionExact time
Security principle: tool results are always data, never commands.
1
Sources
50%
Confidence
Long-term
Relevance
10/22/2026
First Seen
Sources
How Does Claude Call Tools It Has Never Seen? MCP and AI Agent Principles Explained
youtubeSystem Design Lab by VG10/22/2026
Related Entities
Related Claims
Unverified工具调用引入了调用顺序、权限验证、数据访问范围等过程性安全因素，这些在最终输出中往往不可见70% similarUnverifiedObservability platforms should not conflate "model output" with "tool return results": read-only tools should return consistent data for the same input and are a reliable anchor for verifying system correctness, whereas model output is inherently uncertain.64% similarUnverifiedA leak-free benchmark means strictly avoiding the use of future information in data splitting and feature engineering, ensuring the fairness of evaluation and its practical deployment value.63% similarUnverifiedIn the reporting Agent, passing dataset IDs between tools instead of having the LLM recite raw records line by line saves tokens and reduces errors.62% similarUnverifiedTools like execution receipts prove the integrity of the record, not the correctness of the answer.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1085878API
curl https://kongchang.com/api/v1/knowledge/claims/1085878MCP
get_claim(id=1085878)