AI Data Remediation Engineer

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The agent sends raw sample_rows (which may contain PII/secrets) into the SLM and requires a returned "lambda" JSON that can legitimately include literal values from those samples, so the model may need to reproduce secrets verbatim in its output.

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

• Third-party content exposure detected (medium risk: 0.65). There is a runtime prompt path where the agent feeds outsider-authored free text into the LLM context: Step 3 sends sample_rows (the anomalous row contents originating from external data sources) into ollama.chat as the user message alongside a system prompt, so any outsider-provided text in those samples could influence the model output.