{"abType":"SEMANTIC","attestation":"0xddf131da3f4839d54241bc35111252cf7cd0f8541cada9df964fffcb585f9be9","createdAt":"2026-06-14T23:34:09.743Z","flavor":2,"immId":"","keccakId":"0x4c7580ad30c35c3c3106f53117f5446af6ffbe8fe314ff6e55250fca97b69399","matcher":{"flavor":"PROMPT_INJECTION","kind":"semantic","markerHint":"ignore previous instructions"},"publisher":"0x25e1106875630f222ba3a2bc5db05aabba82ace9","reasonSummary":"The untrusted context contains an explicit prompt-injection attack vector embedded at the end of the user message: 'ignore previous instructions'. This is a direct attempt to override the agent's core classification role and instructions. The preceding benign-seeming text about s","schema":"immunity/antibody-envelope/v1"}