hunt-rag-vector

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This file is an explicit offensive playbook describing persistent corpus poisoning, cross-tenant data theft, and OOB exfiltration techniques—clear malicious intent to steal or leak data.

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

• Third-party content exposure detected (high risk: 0.85). This skill’s runtime workflow is about querying/searching a vector database and then feeding retrieved “sources”/similar chunk text (and possibly IDs/metadata returned alongside embeddings) into the agent’s LLM, which is outsider-authored free text when the system indexes other tenants/users’ documents or web/third-party uploads.