screenshot
Pass
Audited by Gen Agent Trust Hub on Mar 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
cleanshotcommand to perform various desktop operations like capturing the screen, recording video, and toggling desktop icons. - [PROMPT_INJECTION]: The
capture-textsubcommand performs OCR on screen regions or files, creating a vulnerability surface for indirect prompt injection. Malicious text visible on a user's screen could be ingested and interpreted as instructions by the agent. - Ingestion points: Text content extracted from the screen or image files via the
capture-textcommand. - Boundary markers: None.
- Capability inventory: Execution of shell commands via the Bash tool as specified in the skill configuration.
- Sanitization: There is no evidence of sanitization or filtering applied to the text results returned from OCR before the agent processes them.
Audit Metadata