antigravity-2
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill frequently executes a local PowerShell script (antigravity.ps1) using the -ExecutionPolicy Bypass flag to run unsigned code.
- [PRIVILEGE_ESCALATION]: The use of -ExecutionPolicy Bypass is a technique to circumvent system-level execution restrictions on PowerShell scripts.
- [DATA_EXFILTRATION]: The skill accesses sensitive local application data to establish a connection, including reading CSRF tokens from the command-line arguments of the language_server.exe process and retrieving the DevTools port from a file in the user's %APPDATA% directory.
- [INDIRECT_PROMPT_INJECTION]: The skill functions as a bridge that transmits untrusted user-provided instructions and goals to a local desktop application.
- Ingestion points: User-provided goal, workspace, and nextStep instructions (SKILL.md).
- Boundary markers: The skill mentions that the agent should perform safety checks but does not define strict delimiters for the user-supplied content.
- Capability inventory: Ability to execute PowerShell scripts, write files to the bridge job folder, and drive application UI via DevTools (SKILL.md).
- Sanitization: There is no evidence of input validation or sanitization for the user-provided goals before they are processed by the bridge or injected into the app.
Audit Metadata