The Agent Skills Directory

[SAFE]: The skill is purely instructional and defines a structured workflow for image generation using specific MCP tools. It does not contain any malicious scripts, commands, or network operations.
[SAFE]: Security guardrails are explicitly defined for file system access, instructing the agent to keep the output_path within a designated local directory (image_out) and to avoid absolute paths, which prevents directory traversal risks.
[SAFE]: The skill includes instructions to sanitize potentially unsafe or irrelevant input by extracting patterns and rewriting them into safe templates, reducing the risk of indirect prompt injection from source materials.

nanobanana