skills/manaflow-ai/cmux/cmux-cloud-vm/Gen Agent Trust Hub

cmux-cloud-vm

Pass

Audited by Gen Agent Trust Hub on Oct 29, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from remote terminal outputs, creating an attack surface for indirect prompt injection.
  • Ingestion points: Data is ingested via cmux vm terminal read, cmux vm terminal wait, and cmux vm terminal output as described in references/commands.md.
  • Boundary markers: No specific delimiters or instructions are provided to the agent to treat output from the remote machine as untrusted content.
  • Capability inventory: The skill provides powerful capabilities including arbitrary code execution (cmux vm exec), network policy modification (cmux vm network), and service publishing (cmux cloud domains publish).
  • Sanitization: Beyond regex matching for specific triggers in terminal output, there is no evidence of input validation or sanitization.
  • [COMMAND_EXECUTION]: The skill executes commands both locally and remotely as part of its core functionality.
  • Local Execution: Commands such as cmux vpn up require elevated privileges (sudo) to manage WireGuard network interfaces on the host Mac.
  • [REMOTE_CODE_EXECUTION]: The primary purpose of the skill is to facilitate remote code execution on cloud VMs via cmux vm exec and cmux vm run. It also supports automated agent updates sourced from official GitHub release assets as documented in references/commands.md.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 29, 2026, 11:41 AM