The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands using the muapi CLI and curl. Specifically, it directs the agent to substitute user-provided inputs into command strings, which creates a command injection surface if the inputs contain shell metacharacters.
[EXTERNAL_DOWNLOADS]: The skill communicates with an external API at api.muapi.ai to perform its core image processing functions. This involves sending image URLs and processing instructions to a remote server.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it interpolates untrusted user data into its operational prompts without sanitization or boundary markers.
Ingestion points: Inputs product_name, model_description, and setting from SKILL.md are used directly in prompts.
Boundary markers: None provided for user-supplied variables.
Capability inventory: Access to muapi CLI tools and curl for network operations.
Sanitization: No evidence of validation or escaping for the interpolated strings.
[DATA_EXFILTRATION]: The instructions demonstrate how to pass the $MUAPI_API_KEY environment variable within a curl command. While this is part of the intended service interaction, exposing credentials in a shell execution context where user-controlled strings are interpolated increases the risk of accidental exposure or exfiltration.

muapi-ugc-lifestyle-try-on