The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input from the premise and style parameters and interpolates them directly into image generation prompts, creating a surface for indirect prompt injection. 1. Ingestion points: SKILL.md (premise, style inputs). 2. Boundary markers: The prompts use double quotes for the beat description to provide some delimitation. 3. Capability inventory: Execution of muapi CLI tools and curl network requests via shell. 4. Sanitization: No explicit validation or filtering of input content is implemented.
[COMMAND_EXECUTION]: The execution instructions direct the agent to construct shell commands using the muapi CLI and curl utility, substituting user-supplied variables and the MUAPI_API_KEY environment variable. This pattern presents a potential command injection risk if the agent does not properly escape shell characters in the premise or style inputs.

muapi-storyboard