screenpipe
Pass
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill facilitates searching for sensitive information across the system. The documentation provides a specific example for searching for "api key" within captured screen data.
- [COMMAND_EXECUTION]: The skill interacts with the local system via the
screenpipeCLI for data retrieval and daemon management, and usescurlto query a local REST API on port 3030. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection from captured data. Ingestion points: OCR and audio transcriptions are retrieved from the user's environment. Boundary markers: No delimiters or safety warnings are present to distinguish captured content from system instructions. Capability inventory: The agent can access a comprehensive history of screen activity and audio. Sanitization: No filtering or sanitization of the retrieved text is performed.
- [EXTERNAL_DOWNLOADS]: Installation instructions point to external resources including
https://screenpi.pe/onboardingand a source repository athttps://github.com/mediar-ai/screenpipe.
Audit Metadata