routeros-qemu-chr

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous patterns for launching virtual machines using qemu-system-x86_64 and qemu-system-aarch64. It also involves the use of qemu-img, sshpass, scp, wget, and unzip for image preparation and VM management. These utilities are used within shell scripts and programmatic wrappers (e.g., using Bun.spawn).
  • [EXTERNAL_DOWNLOADS]: Information and disk images are downloaded from official MikroTik domains including download.mikrotik.com, upgrade.mikrotik.com, and cdn.mikrotik.com. It also references the author's supplementary tools from GitHub repositories under the tikoci organization, which are utilized for UEFI-bootable images (fat-chr) and automation frameworks (quickchr).
  • [PRIVILEGE_ESCALATION]: In the documentation for GitHub Actions (references/github-actions-ci.md), the skill suggests modifying system-wide udev configurations. Specifically, it provides a command to create /etc/udev/rules.d/99-kvm4all.rules with the MODE="0666" option. This makes the /dev/kvm device node world-writable to enable hardware acceleration for the non-root runner user on ephemeral CI VMs.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect injection surface because it ingests data from external sources and the VM's own API to control subsequent automation steps.
  • Ingestion points: Fetching the latest version from upgrade.mikrotik.com and performing health checks via HTTP/REST on the guest VM (port 80).
  • Boundary markers: No explicit delimiters or instructions are used to ignore content within the ingested data.
  • Capability inventory: Extensive command execution capabilities (QEMU, SSH, SCP) and network operations (CLI-based and language-native fetches).
  • Sanitization: Minimal sanitization (trimming) is applied to the version string before it is used to construct download URLs.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 07:02 AM