The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to interact with the ovpackage CLI tool through shell commands. The instructions correctly advise the agent to locate the binary on the PATH or within a local virtual environment before execution.
[CREDENTIALS_UNSAFE]: The documentation references authentication methods including API tokens and cloud credentials. However, it follows security best practices by instructing the agent to use environment variables (OVPACKAGE_AUTH, AWS_ACCESS_KEY_ID, etc.) or existing configuration files rather than hardcoding sensitive data. It correctly warns against interactive authentication in automated environments.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (catalogs, package metadata files, and repository contents). The documentation provides clear boundary markers and specific flags (--progress-report off) to ensure that CLI output is clean and interpreted correctly by the agent, reducing the risk of accidental instruction obedience from processed files.
[SAFE]: All external URLs and references (AWS, Azure, Google Cloud, and NVIDIA Omniverse domains) are pointed toward well-known service providers or official vendor documentation relevant to the tool's core functionality. No suspicious third-party downloads or remote code execution patterns were detected.

ovpackage-cli