harmonyos-device-automation

Pass

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill executes npx @midscene/harmony@1, which fetches and runs code from the npm registry during execution. This is a standard but external dependency mechanism.\n- [COMMAND_EXECUTION]: The skill relies on the hdc (HarmonyOS Device Connector) utility to perform shell commands, launch applications, and manage device state on connected hardware or emulators.\n- [PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection due to its vision-based automation approach.\n
  • Ingestion points: The skill ingests untrusted data in the form of device screenshots via the take_screenshot and act commands.\n
  • Boundary markers: No delimiters or instructions are used to prevent the AI from following commands that might be visible on the device screen.\n
  • Capability inventory: Capabilities include executing shell commands on the mobile device, running remote packages via npx, and performing various UI interactions.\n
  • Sanitization: Visual content is not sanitized before being interpreted by the AI model, which could lead to the agent obeying instructions displayed within a captured application screen.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 9, 2026, 07:22 AM