hz-quest-verify-first

Pass

Audited by Gen Agent Trust Hub on Sep 25, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to retrieve and act upon data from external documentation and connected device outputs, which presents a surface where untrusted or attacker-controlled data could influence the agent's behavior.\n
  • Ingestion points: Documentation pages fetched from Meta developer portals and outputs from headset management commands (e.g., app list, device info, files list).\n
  • Boundary markers: The instructions suggest using citations and recency notes when presenting verified data, providing some context but no strict technical isolation.\n
  • Capability inventory: The skill is granted shell access through Bash (metavr, npx, hzdb) and can manage files (push, pull, remove) on connected VR devices.\n
  • Sanitization: There are no defined mechanisms for the agent to sanitize or filter content retrieved from documentation URLs or tool outputs.\n- [EXTERNAL_DOWNLOADS]: The skill uses npx to fetch the metavr package from the npm registry at runtime to ensure the latest version is used.\n
  • Evidence: Multiple instructions specify the use of npx -y metavr <args> for documentation searches and device management.\n- [REMOTE_CODE_EXECUTION]: Executing the metavr tool via npx involves running code downloaded from a remote registry (npm) during the skill's execution flow.\n- [COMMAND_EXECUTION]: The skill executes various system and device-level commands using the metavr CLI, including querying device state, managing installed applications, and performing file system operations on connected Android-based VR headsets.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 25, 2026, 06:57 PM
Security Audit — agent-trust-hub — hz-quest-verify-first