huggingface-lora-space-builder
Pass
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- Hugging Face Token Management: The skill requests a Hugging Face access token with write scope to access private repositories and publish the generated Space.
- Context: The token is stored using the official
add_space_secretAPI, which is the standard secure method for managing credentials on the Hugging Face platform. It also checks for existing authenticated sessions (get_token()) to minimize unnecessary user prompting. - External Content Processing: The skill ingests data from external model cards to determine inference parameters and UI design.
- Context: While processing external data can introduce risks of indirect prompt injection, the skill mitigates this by presenting the generated
app.py,requirements.txt, andREADME.mdto the user for explicit approval before publication. This provides an important security checkpoint. - Vendor-Owned and Well-Known Dependencies: The skill identifies and installs dependencies from official sources.
- Context: It references official repositories from Hugging Face and well-known research organizations like Lightricks. These downloads are performed as part of the intended installation of the machine learning stack required for the Gradio demo.
- Automated Input Refinement: The skill may automatically prepend trigger words to user prompts if they are defined in the LoRA's model card.
- Context: This is a functional feature designed to ensure the LoRA activates correctly. While it modifies user input, it does so based on the specific requirements of the model being demonstrated.
Audit Metadata