skills/vercel-labs/emulate/google/Gen Agent Trust Hub

google

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • Command Execution and External Downloads: The skill uses npx to run the emulate utility. This is a standard development practice that downloads and executes packages from the npm registry to set up a local testing environment.
  • Local Network Interaction: The skill is configured to interact with a service hosted on localhost:4002. This local-first approach is a security best practice for development, as it ensures that sensitive authentication flows and API calls remain within the developer's controlled environment.
  • Indirect Prompt Injection Surface: The emulator processes mock data, such as email bodies, calendar event summaries, and file contents. This constitutes a potential surface where external data could contain instructions for the agent. The risk is considered low, as this is a standard capability for tools processing user-defined or mock content.
  • Mock Credential Management: The skill provides examples of client secrets and IDs (e.g., GOCSPX-secret). These are clearly identified as mock values for local emulation and do not involve the use or exposure of real production credentials.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 09:59 AM
Security Audit — agent-trust-hub — google