The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands using the muapi CLI tool to perform image editing tasks (e.g., muapi image edit).
[EXTERNAL_DOWNLOADS]: The skill uses curl to interact with the external API at https://api.muapi.ai/api/v1/. This is described as a fallback mechanism for features not yet available in the CLI and is used to send image generation requests and poll for results.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection as it ingests untrusted user inputs (product_name, model_description, setting) and interpolates them directly into prompts for image generation models.
Ingestion points: User-provided inputs defined in the Inputs section of SKILL.md (e.g., product_name, setting).
Boundary markers: None. The inputs are directly embedded into natural language prompts.
Capability inventory: The skill has the capability to execute shell commands (muapi CLI) and perform network operations (curl).
Sanitization: No explicit sanitization or validation of user-provided strings is performed before interpolation.

muapi-ugc-lifestyle-try-on