antigravity-manager

Fail

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONPRIVILEGE_ESCALATIONDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides instructions to download and execute scripts directly in the system shell from a non-trusted repository (github.com/lbjlaq).
  • Evidence includes: curl -fsSL https://raw.githubusercontent.com/lbjlaq/Antigravity-Manager/v4.1.30/install.sh | bash and irm https://raw.githubusercontent.com/lbjlaq/Antigravity-Manager/main/install.ps1 | iex.
  • It also includes remote execution for development tools from well-known services: curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh.
  • [PRIVILEGE_ESCALATION]: The build instructions require the user to run commands with administrative privileges using sudo to install system dependencies.
  • Evidence: sudo apt install libwebkit2gtk-4.1-dev libayatana-appindicator3-dev librsvg2-dev.
  • [DATA_EXFILTRATION]: The skill instructs the agent to access and display sensitive credential files stored on the local filesystem.
  • Evidence: grep -E '"api_key"|"admin_password"' ~/.antigravity_tools/gui_config.json and docker logs antigravity-manager (which is stated to contain recovered keys).
  • [COMMAND_EXECUTION]: The skill encourages extensive use of shell commands to manage Docker containers, install packages, and configure system settings.
  • Evidence: docker run, cargo install tauri-cli, and various npm commands.
  • [INDIRECT_PROMPT_INJECTION]: As the skill operates as an API proxy gateway between the agent and multiple upstream providers, it creates a surface where external model responses could potentially influence the agent's behavior via the proxy pipeline.
  • Ingestion points: The proxy server ingests all prompt and response data for Google and Anthropic accounts.
  • Boundary markers: No explicit boundary markers or instruction-ignoring delimiters are defined for the proxying layer.
  • Capability inventory: The skill has the ability to execute shell commands, read local files, and perform network operations.
  • Sanitization: There is no evidence of sanitization or filtering of the proxied content before it reaches the agent.
Recommendations
  • HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/lbjlaq/Antigravity-Manager/v4.1.30/install.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Sep 12, 2026, 08:37 PM
Security Audit — agent-trust-hub — antigravity-manager