Analysis · curated 15 Jul 2026
Mitigate indirect prompt injection risks from Google Cloud MCP | Google Cloud Data Agent Kit extension for Antigravity IDE | Google Cloud Documentation
First reported · updated · 3 reports google.com
Coverage timeline
Why it matters
Coding agents run with the full delegated privileges of their user, so indirect prompt injection through MCP-accessed data can drive them to compromise infrastructure and exfiltrate data, making these vendor-recommended guardrails important for defenders deploying AI coding agents.
Google Cloud documentation for the Data Agent Kit extension (VS Code and Antigravity IDE) warns that coding agents connected via Google Cloud MCP can be hijacked through indirect prompt injection, where malicious instructions hidden in data sources such as Cloud Storage, BigQuery, email, or calendars are interpreted as commands. The guidance recommends mitigations including running agents in constrained environments like Cloud Workstations with disabled internet access and no root privileges.