The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/generate-nano-art.sh executes an external core script core/media/generate-image.sh using the bash command. The script relies on a dependency located outside the skill's specific directory (../../../../core/media/), which assumes a specific filesystem structure and delegates execution to a script whose logic is not contained within the skill.
[INDIRECT_PROMPT_INJECTION]: The skill takes multiple user-controlled inputs via command-line arguments (such as --subject, --action, --context, and --text) and interpolates them directly into a logic-driven reasoning brief.
Ingestion points: User-provided parameters for the generate-nano-art.sh script, which are likely provided by an AI agent interpreting user requests.
Boundary markers: The constructed prompt uses markers like [REASONING_BRIEF] and [EXECUTE] to structure the brief, but these act as formatting for the image model rather than security boundaries to prevent instruction leakage from the parameters themselves.
Capability inventory: The skill has the capability to execute shell commands (via the bash wrapper) and interface with external image generation primitives.
Sanitization: The script performs no validation, escaping, or filtering of the user-provided strings before they are interpolated into the prompt and passed to the shell environment.

muapi-nano-banana