huggingface-gradio
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references documentation and source code from Hugging Face's official domains and repositories (e.g., gradio.app, github.com/huggingface), which are recognized as trusted organizations.
- [DYNAMIC_EXECUTION]: Documentation for the gr.HTML component (SKILL.md) includes the js_on_load parameter, which allows for client-side JavaScript execution to enable interactivity in custom web components.
- [COMMAND_EXECUTION]: The skill documents the use of the gradio CLI (SKILL.md), specifically the info and predict commands, which are standard tools for programmatically interacting with Gradio applications and remote machine learning models.
- [INDIRECT_PROMPT_INJECTION]: Web interface components described in SKILL.md and examples.md, such as Textbox and Chatbot, serve as ingestion points for untrusted user data. The documentation notes that components like Chatbot and Markdown include built-in sanitization (sanitize_html=True) by default. The primary capability associated with this surface is model inference via network requests.
- [DATA_EXFILTRATION]: The gradio predict CLI command facilitates the transmission of input data to remote Hugging Face Spaces for model processing, representing the intended network functionality of the library.
Audit Metadata