google-slides

Pass

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill performs legitimate automation of Google Workspace tasks using authorized MCP connectors and tools belonging to the vendor's environment.
  • [EXTERNAL_DOWNLOADS]: The skill references documentation and asset locations on official Google developer domains (e.g., developers.google.com, drive.google.com) for API reference and functionality.
  • [DATA_EXFILTRATION]: The skill provides instructions to set specific Google Drive file permissions to 'anyone' with a 'reader' role for generated images. This configuration is a documented functional requirement that allows the Google Slides API to fetch and insert these images into a presentation.
  • [PROMPT_INJECTION]: The skill processes external data from user-provided content briefs and Slide templates, which represents an indirect prompt injection surface.
  • Ingestion points: User-supplied content briefs and existing slide content retrieved via the google_slides_get_slide tool.
  • Boundary markers: The skill relies on specific text tokens (e.g., [Title], [Body]) and YAML-based structural recipes to define the scope of content replacement.
  • Capability inventory: The skill uses tools to duplicate slides, replace text, and modify file permissions within the user's connected Google account.
  • Sanitization: The workflow incorporates a mandatory QA step using a separate review agent to verify that placeholders are correctly substituted and to prevent the unintended retention of template content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 23, 2026, 01:11 PM
Security Audit — agent-trust-hub — google-slides