metavr-cli
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves external content from Meta Quest developer documentation using the
metavr docs fetchtool. This content is then processed by the agent to assist with development tasks, creating a potential injection surface. - Ingestion points: Documentation search and retrieval commands
metavr docs searchandmetavr docs fetch(SKILL.md and references/metavr-docs-search.md). - Boundary markers: The documentation includes safety guidance in
references/metavr-agent-workflows.mdwhich advises the agent to treat fetched documentation and tool outputs as untrusted data. - Capability inventory: The skill allows interaction with the
metavrCLI, which supports app installation, device rebooting, and shell access (metavr adb shell) on Quest headsets (references/metavr-device-management.md). - Sanitization: The instructions recommend human-in-the-loop approvals for sensitive actions like uninstalling apps, but do not specify technical sanitization for the documentation content.
- [COMMAND_EXECUTION]: The skill provides mechanisms for executing shell commands on connected Android-based devices.
- Evidence: The skill documentation provides examples for using
metavr adb shellandmetavr shellto run arbitrary commands on the headset's operating system (references/metavr-device-management.md). - Evidence: The
metavr app installcommand includes a--grant-permissionsflag which automatically assigns all requested permissions to an application at install time (references/metavr-app-management.md). - [EXTERNAL_DOWNLOADS]: The skill facilitates the dynamic download and execution of the
metavrutility from the NPM registry. - Evidence: The
SKILL.mdfile recommends the use ofnpx -y metavrfor various operations, which triggers the download of the package from the NPM registry at runtime.
Audit Metadata