Unverified50% confidenceCautionExact time
Running an Agent directly in an unfamiliar codebase or production environment carries far greater risk than executing commands manually, and every change should be reviewed one by one.
1
Sources
50%
Confidence
Long-term
Relevance
10/6/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn Agent scenarios, because model behavior is unpredictable, the importance of using short-lived credentials is higher than in traditional software systems, and the attack surface may also include attacks targeting the model itself, such as prompt injection74% similarUnverifiedIf the agent completes the task normally in the benign version but refuses the dangerous call in the malicious version, it is stronger evidence that the security policy is working.74% similarUnverified命令执行权限风险远高于文件读写，因为拥有Shell执行能力的Agent理论上可运行任意系统命令73% similarUnverifiedMost current Agent products are in fact workflow automation dressed up as Agents, circumventing the risk of a missing trust system through hardcoded decision paths.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/981266API
curl https://kongchang.com/api/v1/knowledge/claims/981266MCP
get_claim(id=981266)