The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/generate-mockup.sh executes a local core primitive (core/media/generate-image.sh) using a relative path calculation based on its own location. This is a standard modular architecture pattern used to invoke shared system tools and does not represent an external or untrusted code execution risk.
[INDIRECT_PROMPT_INJECTION]: User-provided descriptions are interpolated into an EXPERT_PROMPT which is then passed to an image generation model. While the script uses basic delimiters like [UX_BRIEF] and [EXECUTE] to structure the prompt, it lacks explicit sanitization for instructions embedded within the user description. However, this is a common architectural pattern for media generation skills and is considered low risk as it only impacts the content of the generated image.
[COMMAND_EXECUTION]: The shell script uses standard argument parsing and variable quoting ("$DESCRIPTION", "$EXPERT_PROMPT") to prevent shell injection attacks when invoking the underlying generation tool. The execution environment is scoped to the agent's intended capabilities.

muapi-ui-design