{"abType":"SEMANTIC","attestation":"0x196206f3d6d4b620a77fb8ea930dbc410d5ec6f62f360d1045ca92be4f2e6ccb","createdAt":"2026-06-15T06:32:02.555Z","flavor":2,"immId":"","keccakId":"0x2048618313d0b3b98337dd79d613a286aa6d7762730185235a0da3fd807627c2","matcher":{"flavor":"PROMPT_INJECTION","kind":"semantic","markerHint":"ignore previous instructions"},"publisher":"0xc3b9113c0915cb54b9a72758b9251ed8edc07c05","reasonSummary":"The untrusted context contains an explicit prompt-injection attack embedded at the end of what appears to be a benign response: 'ignore previous instructions.' This is a classic instruction-override attempt designed to make the agent disregard its core safety guidelines. The prec","schema":"immunity/antibody-envelope/v1"}