Unverified50% confidenceSolutionExact time
The fundamental logic of defending against indirect injection is the separation of data and instructions: retrieved content should be confined to a read-only data zone, and hierarchical isolation can be established by placing external content in the tool role rather than the user role.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPrompt injection can be divided into two types: direct injection and indirect injection, where indirect injection refers to attackers hiding malicious instructions in external content read by the agent.70% similarUnverifiedWhen using orchestration frameworks such as LangGraph and involving file-reading or data-reading operations, permission management logic must be explicitly defined at the code level.69% similarUnverifiedWrapping retrieved text in clearly delimited, labeled blocks and explicitly telling the model that the content within those blocks is data rather than instructions can eliminate the majority of the prompt injection attack surface.65% similarUnverifiedIf a piece of content only needs to be read without triggering any logic execution, it should be a Resource64% similarUnverifiedThe read-then-write combination, in which an external write is triggered immediately after reading private data, is the classic path for data exfiltration and should be governed as an independent policy trigger condition.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/998092API
curl https://kongchang.com/api/v1/knowledge/claims/998092MCP
get_claim(id=998092)