doca-rdmi
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user queries about DOCA RDMI programming to drive its workflows and sample modifications. No explicit boundary markers or sanitization logic are defined in the instructions. Ingestion occurs during interactive task guidance in SKILL.md and TASKS.md. Boundary markers are absent. The capability inventory includes environment inspection tools such as pkg-config and ls. Sanitization is not specified.
- [COMMAND_EXECUTION]: The skill facilitates environment setup and debugging using standard system commands. Evidence includes the use of pkg-config and doca_caps for dependency and capability checking in SKILL.md and TASKS.md. Debugging steps in TASKS.md also recommend user-level execution of hardware diagnostic tools like dmesg and mlxconfig.
- [EXTERNAL_DOWNLOADS]: The skill references official vendor documentation and repositories. Evidence includes links to nvidia.com and github.com/NVIDIA-DOCA in the skill metadata files. These references target the official infrastructure of the skill's author and are documented neutrally.
Audit Metadata