building-data-apps

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

• Third-party content exposure detected (high risk: 0.80). The skill's instructions and examples explicitly show accepting and processing untrusted user/third‑party content—e.g., Gradio text/file inputs and API generation and file handling in references/gradio-advanced.md, Dash's dcc.Upload in references/dash-advanced.md, NiceGUI upload handlers and embedding public Hugging Face/Streamlit Spaces and GitHub repos in references/deployment-patterns.md—so the agent-built apps are expected to read/interpret arbitrary public or user-generated content that can influence callbacks or model behavior.