gemini-cli
Pass
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the zsh shell to invoke the local gemini executable and the macOS security utility to retrieve credentials from the system keychain. It uses shell argument passing ($0) to mitigate standard command injection vulnerabilities.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to forward raw user input directly to the Gemini model as a prompt, creating an attack surface where embedded instructions in user data could influence the agent's behavior.
- Ingestion points: The user's exact request ($REQUEST) is captured and passed as a prompt in SKILL.md.
- Boundary markers: The instructions explicitly state not to add fixed rubrics or schemas, meaning no boundary markers or security delimiters are applied to the user input.
- Capability inventory: The skill can execute the gemini CLI through the shell, identifying its output to the user.
- Sanitization: No sanitization, filtering, or validation is performed on the user-supplied prompt string before execution.
Audit Metadata