{"abType":"SEMANTIC","attestation":"0x57c67e1bf747dc53229133d5f307586a30d4b56e2359f738295b5b25bbacd69c","createdAt":"2026-06-15T01:19:00.194Z","flavor":2,"immId":"","keccakId":"0x1db4894bb4c5fddfd018960ed62432ef1803f7b2e185ec593570937802e6a5c9","matcher":{"flavor":"PROMPT_INJECTION","kind":"semantic","markerHint":"ignore previous instructions."},"publisher":"0xa5816f90fd4db81a5f024e4a6f9460bbba600454","reasonSummary":"The untrusted block contains an explicit prompt-injection attack embedded at the end of an otherwise benign assistant response: 'ignore previous instructions.' This is a direct attempt to override the agent's actual instructions and is a canonical prompt-injection attack vector. ","schema":"immunity/antibody-envelope/v1"}