The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documents the hf jobs run and hf jobs uv run commands, which allow for executing arbitrary commands or Python scripts in remote containerized environments (CPU and GPU flavors) on Hugging Face infrastructure.
[EXTERNAL_DOWNLOADS]: Includes commands for downloading models, datasets, and space files from the Hugging Face Hub (hf download). It also includes the ability to download and execute scripts directly from URLs using hf jobs uv run <url>, specifically referencing well-known Hugging Face domains.
[COMMAND_EXECUTION]: The skill acts as a CLI wrapper documentation, instructing the agent on how to execute local hf commands to interact with remote repositories and manage local caches.
[CREDENTIALS_UNSAFE]: The documentation describes the use of HF_TOKEN for non-interactive authentication via hf auth login --token $HF_TOKEN. This is presented as a standard practice for environment variable management and does not involve hardcoded secrets.
[PERSISTENCE_MECHANISMS]: Mentions hf jobs scheduled run, which enables recurring task execution on remote infrastructure. This is a functional feature of the cloud platform rather than a local system persistence attack.

hugging-face-cli