The Agent Skills Directory

COMMAND_EXECUTION (LOW): The skill grants access to the gemini CLI via a broad Bash allowance (gemini *). While intended for development assistance, this tool possesses the capability to modify files and execute local system commands.
Indirect Prompt Injection (LOW): The skill is vulnerable to indirect prompt injection attacks because it processes external data through the Gemini CLI without sanitization.
Ingestion points: Positional prompt arguments in SKILL.md are passed directly to the gemini command.
Boundary markers: Absent. There are no delimiters or instructions to ignore embedded commands within the input string.
Capability inventory: The skill allows full use of the gemini CLI, which supports tool execution and local file access.
Sanitization: Absent. Input is interpolated into a shell command without escaping or validation.
EXTERNAL_DOWNLOADS (SAFE): The skill requires the gemini CLI to be pre-installed. The official google-gemini/gemini-cli is a trusted source, and the skill does not perform any unverified downloads at runtime.

ask-gemini