Unverified50% confidenceSolutionExact time
Narrowing permissions to the minimum set required for the task (such as read-only, limited to specific repositories) means that even if the model is persuaded by an injection, it will be rejected at the tool-call layer.
1
Sources
50%
Confidence
Long-term
Relevance
10/22/2026
First Seen
Sources
Can AI Agents Be Socially Engineered? Understanding Prompt Injection and Tool Permission Risks
youtubeRuleArena 攻防實驗室10/22/2026
Related Entities
Related Claims
Verified最小权限原则要求任何程序或用户只拥有完成当前任务所必需的最低权限78% similarUnverifiedThe principle of least privilege requires that each task be assigned exactly the permissions needed during the agent design phase, rather than opening up a broad set of capabilities.77% similarUnverifiedThe principle of least privilege has an often-overlooked precondition: the permissions granted must be sufficient to complete the task; when permissions are drawn too narrowly, the subject cannot perform its duties, and in practice people look for workarounds, which instead creates more covert and harder-to-audit temporary channels.73% similarUnverifiedThe industry-standard approach to addressing data sensitivity is the "principle of least privilege," whereby each integrating party can access only the minimal dataset necessary to complete a specific task.72% similarUnverifiedThe tools available to an agent should be strictly limited to only the functionality it truly needs (least-privilege tool invocation).72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1085885API
curl https://kongchang.com/api/v1/knowledge/claims/1085885MCP
get_claim(id=1085885)