Unverified50% confidenceDecision RuleExact time
Token input/output should be tracked per agent rather than only per request, in order to identify which node is causing problems.
1
Sources
50%
Confidence
Long-term
Relevance
10/5/2026
First Seen
Sources
Related Entities
Related Claims
Unverified攻击者通过提示词注入控制Agent的决策逻辑后，可利用Agent已获授权的所有工具和接口实现全链路攻击，形成混淆代理人（Confused Deputy）问题69% similarUnverifiedEmbedded agents fundamentally rely on the host product's API completeness; if a feature has no corresponding callable interface, the agent cannot execute on behalf of the user.66% similarUnverifiedWhen selecting an agent inference service, attention should be paid to its end-to-end performance under real agent workloads, rather than only looking at decoding speed on benchmarks.66% similarUnverifiedThe identity layer ensures that every call made by every Agent has a verifiable identity attribution, which is a prerequisite for enforcing access control policies.66% similarUnverifiedAssigning a unique identifier to each Agent resolves the risk that multiple Agents sharing the same set of credentials makes it impossible to identify the responsible party.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/976081API
curl https://kongchang.com/api/v1/knowledge/claims/976081MCP
get_claim(id=976081)