ling-gui-agent-skill
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [SAFE]: No malicious patterns detected. The skill follows security best practices for credential management and user safety.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a vision-based attack surface as it processes screenshots of potentially untrusted environments. 1. Ingestion points: User goal text and device screenshots processed in
executor.pyandplanner/ling.py. 2. Boundary markers: Detailed system prompts inplanner/ling.pydefine the interaction model. 3. Capability inventory: Pointer actions and keyboard input viaPyAutoGUIandAppium. 4. Sanitization: Model outputs are parsed using strict regex against allowed functions inLingTemplate.parse_model_output. - [COMMAND_EXECUTION]: The skill uses functional system commands for environmental discovery and logging, including
adbfor Android discovery inappium_client.py,ffmpegfor diagnostic video generation inexecutor.py, and Python import probes inscripts/run.py. - [EXTERNAL_DOWNLOADS]: The skill requires installation of standard, well-known Python dependencies from official registries as specified in
pyproject.tomland communicates with user-configured model endpoints.
Audit Metadata