skills/metamask/skills/gemini/Gen Agent Trust Hub

gemini

Pass

Audited by Gen Agent Trust Hub on Jul 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is designed to ingest and analyze external, untrusted content such as entire codebases and technical documentation (ingestion points). When paired with the Gemini CLI's ability to auto-approve tool execution in 'yolo' mode (capability), this creates a surface for indirect prompt injection. The skill lacks instructions for implementing boundary markers or specific sanitization routines to prevent analyzed content from influencing the agent's behavior (sanitization/boundaries).
  • [COMMAND_EXECUTION]: The skill documents the use of the gemini CLI tool with flags like --approval-mode yolo and -y, which allow for autonomous operation. While the skill mandates that the agent use AskUserQuestion to seek permission before utilizing these high-impact flags, the capability itself grants significant system access.
  • [COMMAND_EXECUTION]: The skill includes instructions for using shell commands such as ps, pkill, kill, and lsof to monitor and terminate Gemini processes. While these are presented as troubleshooting steps for managing hung processes in non-interactive environments, they represent a capability for system process management.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 13, 2026, 06:10 PM
Security Audit — agent-trust-hub — gemini