investigating-ransomware-attack-artifacts

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). The required workflow ingests outsider-authored free text from ransom notes and other evidence files on the target filesystem (e.g., collect_ransom_notes() reads content_preview from matched README*.txt/DECRYPT*.txt/HOW_TO*.txt files), which can contain attacker-controlled prompt-injection text that is then placed into the agent’s LLM context via the generated report/JSON outputs.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs running "sudo insmod lime.ko" (inserting a kernel module) and performs filesystem writes under root paths (e.g., /cases), which require elevated privileges and modify the host system state.