hugging-face-space-deployer
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill attempts to access the Hugging Face authentication token located at
~/.cache/huggingface/tokeninscripts/create_space.py. This is a sensitive file path used to verify authentication status before performing repository operations. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted metadata and configuration data from the Hugging Face Hub to automate deployment decisions and generate code templates.
- Ingestion points: Data is retrieved from remote repositories using
api.list_repo_files,hf_hub_download, andmodel_infowithinscripts/deploy_model.py. - Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious instructions embedded in the remote metadata.
- Capability inventory: The skill has the capability to create new repositories, upload files, and manage platform secrets via
scripts/create_space.pyandscripts/manage_space.py. - Sanitization: The skill performs structural checks for filenames but does not sanitize the content of remote configuration files (like
adapter_config.json) before incorporating them into the deployment workflow.
Audit Metadata