analyzing-malware-behavior-with-cuckoo-sandbox

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

• Third-party content exposure detected (high risk: 0.80). This skill explicitly accepts arbitrary URLs for sandboxing (SKILL.md "Submit URL for analysis" / API examples) and the agent code (scripts/agent.py) loads Cuckoo JSON reports and reads network/http entries (including response bodies) from those reports, which are untrusted, user-provided web content that the agent ingests and uses to compute signatures/scores and summary decisions.