The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions direct the agent to use uvx to download and execute the hf-mem package. This involves fetching and running external code at runtime, which is the primary intended function of the skill.- [COMMAND_EXECUTION]: The skill relies on executing shell commands via uvx hf-mem to perform its tasks. These commands are executed in the host environment where the agent operates.- [EXTERNAL_DOWNLOADS]: The use of uvx triggers downloads of the hf-mem package from the Python Package Index (PyPI).- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through command injection. It suggests building shell commands by interpolating user-provided model IDs and file paths (e.g., --model-id <model-id>). An attacker could provide a model ID containing shell metacharacters to execute arbitrary commands.
Ingestion points: User-supplied model identifiers and GGUF filenames.
Boundary markers: None are defined in the instructions to separate user input from the shell command.
Capability inventory: The skill has the ability to execute shell commands via uvx.
Sanitization: The instructions do not include steps to sanitize or validate the model IDs before execution.

hf-mem