Unverified50% confidenceOpinionExact time
Input filtering alone can hardly provide exhaustive defense against indirect prompt injection, because plausible-looking instructions can be naturally embedded in legitimate documents; a more reliable approach is to treat read content as untrusted and limit, at the tool call boundary, what follow-up actions it can trigger.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThere is no reliable 'detect and block' solution for prompt injection, because it is difficult to judge whether a piece of text is an ordinary description or a malicious instruction, and to the model the two are identical in form; isolation and permission control are more realistic than content filtering.80% similarUnverified用户可以通过提示注入和语义改写等手段绕过内容检测，使实际拦截效果是概率性的而非绝对的76% similarUnverifiedIn adversarial input scenarios, both typed and generative judges read the content being scored, and the content may carry instructions targeting the judge, so written reasoning cannot reliably expose prompt injection.76% similarUnverified折中方案是将是否提供可复现代码作为评审的显式打分项，而非一刀切硬性拒稿，并对确有正当理由无法公开代码的工作给予豁免74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1068608API
curl https://kongchang.com/api/v1/knowledge/claims/1068608MCP
get_claim(id=1068608)