fleet-hunting-with-velociraptor

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/agent.py, the runtime can ingest outsider-provided free text via the query subcommand argument (vql) and the validate subcommand path contents (arbitrary YAML query strings), which are then passed to velociraptor query / parsed at runtime.

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 includes explicit privileged operations that modify system state (e.g., "sudo mv" to /usr/local/bin and instructions to install/run the agent as a service) and thus asks an operator to perform actions requiring root/administrator privileges.