erne-setup-device
Warn
Audited by Gen Agent Trust Hub on Apr 9, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The setup instructions configure an MCP server using 'npx -y agent-device'. The '-y' flag bypasses npm's safety prompt, automatically downloading and executing the 'agent-device' package from the public registry. This allows the execution of code not provided in the skill package itself.
- [EXTERNAL_DOWNLOADS]: Installation involves downloading the 'agent-device' package from the npm registry.
- [COMMAND_EXECUTION]: Executes system-level commands including 'xcodebuild', 'xcrun', 'adb', and 'emulator' to inspect and control the local mobile development environment. It also instructions the agent to modify the '.claude/settings.json' configuration file.
- [PROMPT_INJECTION]: The visual control capabilities (screenshots and UI interaction) enabled by this skill create an attack surface for indirect prompt injection from device content. 1. Ingestion points: Screen captures and UI hierarchy data from simulated devices. 2. Boundary markers: Absent; there are no instructions to treat content from the device as untrusted. 3. Capability inventory: Includes full UI interaction (tap, type, swipe, and navigation). 4. Sanitization: None specified for data captured from the device screen.
Audit Metadata