protection-shield

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

• Third-party content exposure detected (medium risk: 0.30). The recurring review reads named server logs on a schedule, and an outsider can write attacker-controlled text (request paths, User-Agent, auth failure fields) into those logs without the agent selecting any specific item.

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

• Attempt to modify system services in skill instructions detected (low risk: 0.30). The skill is a gated defensive incident-response workflow that can stop services, change firewall rules, and patch systems, but it requires explicit authorization, bounded scope, and rollback, and it forbids admin escalation and broad blocks, so it does not push the agent toward unauthorized or sudo-level compromise.