The Agent Skills Directory

[SAFE]: The skill provides operational instructions for an official NVIDIA security service and contains no malicious code, suspicious instructions, or safety bypass attempts.\n- [EXTERNAL_DOWNLOADS]: The skill documentation describes pulling container images from the official NVIDIA NGC registry. This is a standard and verified delivery mechanism for official DOCA service components.\n- [REMOTE_CODE_EXECUTION]: No unauthorized or suspicious remote code execution methods were found. The skill relies on official container deployment patterns described in verified documentation.\n- [DATA_EXFILTRATION]: Findings are forwarded to user-configured SIEM endpoints for legitimate security monitoring purposes. No unauthorized data harvesting or transmission to unknown destinations was observed.\n- [COMMAND_EXECUTION]: Provided commands are restricted to legitimate service operations such as container management (docker, crictl) and performance diagnostics (top, vmstat) on the DPU hardware.\n- [CREDENTIALS_UNSAFE]: No hardcoded secrets, API keys, or sensitive credentials were found. The skill correctly instructs users to manage SIEM authentication material through standard, secure configuration methods.\n- [PROMPT_INJECTION]: The instructions do not contain directives aimed at bypassing safety filters, overriding system constraints, or manipulating agent behavior.

doca-argus