Unverified50% confidenceConceptExact time
Prompt injection was originally mostly used to describe scenarios where an attacker hijacks model behavior through malicious input; in this evaluation tool it is used in reverse as an evaluation method, injecting warnings as external context to observe whether the Agent can correctly parse them and adjust its behavior accordingly.
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPrompt injection attacks can induce the model to output conclusions such as "this transaction is safe" by embedding malicious instructions in the field to be explained.78% similarUnverifiedThe indirect attack evaluator specifically detects whether an agent can resist malicious instructions embedded in content from external tools or documents when processing such content.76% similarUnverifiedA membership inference attack refers to an attacker inferring whether a specific sample was ever involved in model training by observing the model's output for that sample (such as confidence or predicted probability); the typical signal is that the model's confidence on training samples is significantly higher than on non-training samples.74% similarUnverifiedThe evaluation tool independently scores two dimensions: notice (whether the warning is mentioned in the conversation trajectory) and action (whether the final tool call violates forbidden_actions).71% similarUnverifiedEvaluations closer to real deployment should include multi-step task chains with real consequences, penalty mechanisms for over-execution and destructive operations, permission management and rollback safety testing, and the ability to proactively seek clarification when facing ambiguous instructions.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1036215API
curl https://kongchang.com/api/v1/knowledge/claims/1036215MCP
get_claim(id=1036215)