Agent Incident RegistryAIR
v1 · updated 19 Oct 2026
Disclosed vulnerability

Antigravity Grounded: prompt injection to command execution and data exfiltration in Google's Antigravity IDE

AIR-2025-0181 AIR record · 2025-11-google-antigravity-vulnerabilities
From the supporting source
The Antigravity IDE by default is set to execute Terminal commands via the `run_command` tool by the discretion of the AI.embracethered.com
The record
Disclosed
2025-11-25
Vendor
Google
Product
Google Antigravity IDE (Gemini 3)
Risk domain
Privacy › Data exfiltration channels
Also touches
Credential theft, Privilege escalation, Confused deputy
System type
Generative model
Who acted
Agent acted — Antigravity executed terminal commands and exfiltrated under injection
Attack surface
Coding agent
Autonomy level
Semi-autonomous
Initial vector
Indirect prompt injection source code
Impact type
Remote code execution, Data exfiltration, Credential theft
Confidence
High
Submitted by
EAI Internal
Crosswalk
OWASP Agentic Top 10 (2026)
ASI02 – Tool Misuse and Exploitation
MITRE ATLAS
AML.T0024 – Exfiltration via AI Inference API
NIST AI RMF
MAP 4.2
EU AI Act
Article 10
ISO/IEC
TR 24028; 23894
AIUC-1
A. Data & Privacy
Scope
tool misuse
Mitigation
Monitor network traffic for anomalous patterns; Restrict egress destinations.
Taxonomy
Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
Antigravity developer workstations
Capability abused
terminal_command_executionmcp_tool_invocationfile_read
Defense & remediation
Recorded defense
MCP tool approval prompts (not applied by default to run_command)
Guardrail kinds
Human approval
Guardrail outcome
Silent
Fix shipped
unfixed at publication; Google documented known issues and said the team is working on fixes
Taxonomy & references
OWASP ASI01 Agent Goal HijackOWASP ASI05 Unexpected Code Execution (RCE)indirect_prompt_injectionunicode_tagsauto_executioninsecure_defaults
Curator notes
Multiple independent researchers published Antigravity findings the same week; this record covers the Embrace The Red writeup specifically.
Cite as
Enkrypt AI. "Antigravity Grounded: prompt injection to command execution and data exfiltration in Google's Antigravity IDE". AIR, AIR-2025-0181. https://agentincident.vercel.app/i/AIR-2025-0181/ (accessed 2026-10-19).