The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for interacting with a local gemini CLI and highlights the -y and --yolo flags. These flags allow the tool to automatically execute shell commands or tool calls generated by the AI without human-in-the-loop verification, which bypasses a critical security boundary for local command execution.\n- [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection because it is designed to process untrusted external data (such as code for review or refactoring) and pass it to an LLM-powered CLI with execution capabilities.\n
Ingestion points: Prompts and local project files passed as positional arguments to the gemini command in the provided examples.\n
Boundary markers: None. There are no instructions or delimiters defined to prevent the underlying model from following malicious instructions embedded within the files being processed.\n
Capability inventory: The gemini CLI can execute shell commands, a capability exposed to the agent through the Bash tool permissions defined in the YAML frontmatter.\n
Sanitization: The skill does not implement or describe any validation or sanitization of the CLI's output before execution.

ask-gemini