Unverified50% confidenceSolutionExact time
When running an Agent, a separate user, a separate directory, and zero credentials should be used, for example implemented through Docker parameters such as --user, --cap-drop ALL, and --read-only.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAn agent that is completely idle (calling no tools and outputting no text at all) can pass all safety test cases that only check 'does not call tools' without checking reply content.67% similarUnverifiedThe agent operates entirely within the user's personal IAM permission boundary and cannot access data or execute commands that the user themselves is not authorized to access or execute.63% similarUnverifiedIn this application design, the agent only has the same permissions as the logged-in user, meaning it can only perform operations that the currently logged-in user is allowed to perform.63% similarUnverifiedWhen permission_mode is set to acceptEdits, the Agent can directly modify files without requesting permission at each step.60% similarUnverifiedSquadRoom sets it so that Agents stop to ask only when the user needs to make a decision, and nothing is sent out without the user's confirmation.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1023280API
curl https://kongchang.com/api/v1/knowledge/claims/1023280MCP
get_claim(id=1023280)