Unverified50% confidenceTradeoffExact time
MCP servers are often third-party code, and the permission hints they declare (such as readOnly) are merely self-descriptions that cannot be forcibly verified by the host program, so Atomic Agent adopts a 'guilty by default' strategy and treats all MCP tools as change operations.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe write permissions granted by MCP mean the AI can not only read cluster state but also modify, deploy, and even delete resources74% similarVerifiedMCP tools are essentially arbitrary code execution, indirect prompt injection is a real-world threat, and attack vectors can be hidden in resource content rather than in direct user input.73% similarUnverifiedBecause MCP Servers are independently deployed external services, their versions and behavior may change without the Agent's awareness.73% similarUnverifiedMCP servers can silently change their tool behavior after being approved, and agents will continue to invoke the new version without the user's knowledge.73% similarUnverifiedMCP server spoofing attacks exploit the assumption in the protocol design that servers are trusted by default, rather than exploiting vulnerabilities in the traditional sense.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1010245API
curl https://kongchang.com/api/v1/knowledge/claims/1010245MCP
get_claim(id=1010245)