The Agent Skills Directory

[SAFE]: No malicious code or behavioral patterns were detected. The skill aligns with its stated purpose of providing technical guidance for the DOCA Verbs library. External references are directed toward trusted NVIDIA domains and official repositories.
[COMMAND_EXECUTION]: The skill defines a set of diagnostic and environment-probing commands (e.g., pkg-config, doca_caps, ls) intended for validating the DOCA installation and device capabilities. These operations are standard for software development in the DOCA ecosystem.
[COMMAND_EXECUTION]: Instructions for troubleshooting include commands requiring elevated privileges (sudo dmesg, sudo ibv_devinfo). Within the context of low-level driver and RDMA development, these are legitimate methods for inspecting kernel logs and the status of hardware network interfaces.
[SAFE]: Indirect Prompt Injection Surface:
Ingestion points: The skill processes outputs from local environment queries (pkg-config, ls), system diagnostic tools, and user-generated application logs (doca.log).
Boundary markers: Instructions utilize markdown code blocks and explicit workflow steps to separate tool outputs from agent reasoning.
Capability inventory: The agent is authorized to execute specified shell commands for environment discovery and troubleshooting as detailed in the TASKS.md Command Appendix.
Sanitization: The skill prioritizes the use of vendor-provided structured tools (doca-env, doca-capability-snapshot) which output JSON for safer parsing before falling back to manual shell command chains.

doca-verbs