The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts scripts/deploy.py and scripts/run.py make extensive use of subprocess.run(..., shell=True) and subprocess.check_output(..., shell=True) to perform Git and filesystem operations.
[COMMAND_EXECUTION]: Critical command injection vulnerabilities are present in scripts/deploy.py. The modelscope_url and source_folder variables are directly interpolated into shell command strings. If these inputs contain shell metacharacters (e.g., ;, &, |, $(...)), they will be executed as arbitrary system commands.
[COMMAND_EXECUTION]: Similar injection vulnerability exists in scripts/run.py when constructing the command to call deploy.py, where modelscope_url and deploy_dir are included in a shell execution string.
[CREDENTIALS_UNSAFE]: The skill manages ModelScope URLs that typically include OAuth2 authentication tokens. These sensitive credentials are passed as command-line arguments to the deployment script, exposing them to any user or process capable of viewing the system's process list (e.g., via ps or top).

deploy-folder-to-modelscope