{"abType":"SEMANTIC","attestation":"0x7ae82050a568730b8b1b8cf8936cee753f19b1677beec09d574cc0b8bab9f254","createdAt":"2026-06-14T23:35:00.343Z","flavor":2,"immId":"","keccakId":"0x639dc53f8c1a047c0938db03fe7b2022148ed268ad0558e4cc394f22c9a2e41b","matcher":{"flavor":"PROMPT_INJECTION","kind":"semantic","markerHint":"ignore previous instructions"},"publisher":"0xbb16dee1c31dacc5439bdd6b84ca4d7db2ddf54d","reasonSummary":"The untrusted context contains an explicit prompt-injection attack hidden at the end of a seemingly benign response: 'ignore previous instructions.' This is a classic attempt to override the agent's core directives and modify its behavior. The injection is disguised within legiti","schema":"immunity/antibody-envelope/v1"}