The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted user data via media URL inputs which are then processed by the agent.
Ingestion points: The user_image and reference_video inputs defined in SKILL.md are used as dynamic placeholders.
Boundary markers: No delimiters or safety instructions are provided to the agent to treat these inputs as literal data rather than potential command components.
Capability inventory: The skill directs the agent to execute shell commands using the muapi CLI and curl utility.
Sanitization: The skill lacks instructions for sanitizing or escaping user-provided strings before they are interpolated into the shell execution strings.
[COMMAND_EXECUTION]: The instructions command the agent to perform shell operations (using muapi and curl) that incorporate user-provided variables. This direct substitution model creates a potential vulnerability where a malicious input could break out of the intended command context to execute arbitrary shell commands.
[DATA_EXFILTRATION]: The skill naturally transmits user-provided image/video URLs and an API key (via the MUAPI_API_KEY environment variable) to the external endpoint api.muapi.ai. While this is the intended primary purpose of the skill, it involves the transmission of potentially sensitive data to a third-party service.

muapi-cartoon-dance-animation