dpdk

Pass

Audited by Gen Agent Trust Hub on Jun 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Instructions involve using sudo to modify kernel settings in /sys/kernel/mm/hugepages/, mount hugetlbfs volumes, and load kernel modules (modprobe vfio-pci). These privileged commands are required for DPDK initialization and hardware interaction.
  • [COMMAND_EXECUTION]: Uses dpdk-devbind.py for binding hardware network interfaces to userspace drivers, which alters system hardware configuration.
  • [PROMPT_INJECTION]: The skill defines a workflow for raw network packet ingestion using rte_eth_rx_burst in SKILL.md, establishing an indirect prompt injection surface.
  • Ingestion points: Untrusted data enters the agent context through network packets received via rte_eth_rx_burst in SKILL.md.
  • Boundary markers: The provided code snippets lack delimiters or specific instructions to treat packet data as untrusted content distinct from instructions.
  • Capability inventory: The skill provides instructions for system-level access to kernel parameters and hardware drivers via sudo commands and binary execution.
  • Sanitization: The code samples do not include logic for sanitizing or validating packet contents before processing in the poll-mode loop.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 27, 2026, 12:04 PM