The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands by interpolating user-provided text directly into a string: npx @the-focus-ai/nano-banana "<prompt>". This creates a command injection risk as there are no instructions to sanitize or escape shell-sensitive characters like quotes, semicolons, or backticks that may be present in user-controlled input.
Ingestion points: User requests for image creation and editing in SKILL.md and examples.
Boundary markers: Double quotes used for prompt interpolation are insufficient to prevent escaping.
Capability inventory: Shell execution via npx.
Sanitization: No sanitization instructions or human review checkpoints are provided.
[EXTERNAL_DOWNLOADS]: The skill relies on npx to download and execute the @the-focus-ai/nano-banana package from the npm registry at runtime. While this is a tool provided by the author, runtime execution of remote code introduces a dependency on external registry infrastructure.

nano-banana-imagegen