google-cloud-vision

Pass

Audited by Gen Agent Trust Hub on May 12, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions include the global installation of the @membranehq/cli package from the NPM registry. This package is the official command-line interface provided by the vendor (Membrane) for platform interactions.
  • [COMMAND_EXECUTION]: Various membrane CLI commands are used to manage authentication, establish connections to Google Cloud Vision, and execute image processing actions. These commands are part of the intended integration workflow.
  • [PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it processes data retrieved from Google Cloud Vision (such as OCR text and image labels) which could contain adversarial content designed to influence the agent's behavior. Ingestion points: Data returned from membrane action run (e.g., detect-text, detect-labels). Boundary markers: None explicitly provided in the skill instructions. Capability inventory: The agent can execute CLI commands and manage local configuration through the Membrane toolset. Sanitization: No explicit sanitization or filtering of the Vision API output is mentioned in the skill guidelines.
Audit Metadata
Risk Level
SAFE
Analyzed
May 12, 2026, 02:48 AM
Security Audit — agent-trust-hub — google-cloud-vision