The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references official NVIDIA resources including the NGC container registry (nvcr.io/nvidia/doca/doca) and public GitHub repositories under github.com/NVIDIA. These downloads originate from the vendor's own trusted infrastructure.
[COMMAND_EXECUTION]: The instructions guide the agent to perform standard build and inspection tasks using shell commands like ls, cp, pkg-config, meson, and ninja. These operations are limited to the development environment and are necessary for the skill's primary function.
[INDIRECT_PROMPT_INJECTION]: The skill includes workflows that read and modify local source code files (e.g., meson.build, .c samples). This represents an attack surface for indirect prompt injection from local data. However, the skill implements safety policies, such as deriving from verified samples and using explicit placeholders for user values, to mitigate this risk.
[SAFE]: Analysis of the skill's metadata, instructions, and companions revealed no malicious patterns such as credential exfiltration, persistence mechanisms, or obfuscation. The skill adheres to security best practices for secret management by advising developers to use least-privilege credentials and avoiding the inclusion of secrets in logs or output.

doca-programming-guide