read-line-messages
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
spawnSyncto execute system commands includingscreencaptureandpython3to interact with the macOS Quartz window manager. While these are necessary for the skill's functionality, they represent a broad execution surface for interacting with local applications. - [COMMAND_EXECUTION]: The script dynamically executes a Python snippet via
python3 -cto locate the window ID of the LINE application. This is a common pattern for GUI automation but constitutes dynamic code execution at runtime. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted image data (screenshots of LINE chats) and passes them to a transcription LLM. It includes explicit boundary instructions ('Strict rules', 'Output format') to mitigate the risk of the model obeying instructions found within the chat text, though the risk of indirect injection from chat content remains inherent.
Audit Metadata