skills/nvidia/skills/doca-upgrade/Gen Agent Trust Hub

doca-upgrade

Pass

Audited by Gen Agent Trust Hub on Jul 28, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a robust safety policy that requires explicit user confirmation before any upgrade or downgrade command is executed.
  • [SAFE]: Administrative and hardware-touching actions, such as BlueField BFB reflashing or host package updates, are delegated to specialized skills (doca-hardware-safety, doca-setup) that manage these risks centrally.
  • [SAFE]: All external references, including documentation and code examples, point to trusted, vendor-owned domains and repositories (nvidia.com and github.com/NVIDIA).
  • [PROMPT_INJECTION]: The skill has an attack surface through processing external DOCA release notes to identify upgrade targets (Indirect Prompt Injection).
  • Ingestion points: External release notes accessed via doca-public-knowledge-map.
  • Boundary markers: The workflow requires the agent to report the discovered status to the user and obtain manual confirmation before proceeding.
  • Capability inventory: Tooling for package management (apt, dpkg) and firmware modification (delegated to safety skills).
  • Sanitization: Instructions prevent the agent from asserting release status from internal memory, requiring verification against live documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 28, 2026, 05:10 PM
Security Audit — agent-trust-hub — doca-upgrade