agency-ai-data-remediation-engineer

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The code contains an explicit dynamic-eval of model-generated code with only superficial string checks (allowing many RCE bypasses) and an alert/webhook hook that could be repurposed for exfiltration, creating a high-risk backdoor/remote-exec and data-leak vector despite the stated "air-gapped" intent.

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

• Third-party content exposure detected (high risk: 0.75). The workflow ingests outsider-authored free text via “Receive Anomalous Rows” (Step 1), where suspect_rows are anomalous row contents coming from upstream sources the operating user did not author, and those row strings are then embedded and sent into the LLM context in Step 3 (messages user content includes Samples:\n + sample_rows).