data-quality-analysis

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

• Third-party content exposure detected (high risk: 0.70). The SKILL.md discovery and analysis workflow explicitly requires loading datasets from user-specified sources — e.g., "file path, URL, database table, API endpoint" and "Read the data — Load the file (CSV, Excel, JSON, Parquet, database query result, etc.)" — so the agent will ingest and interpret potentially untrusted third‑party content (arbitrary URLs / API results) that directly drives assessments and report-generation.