The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input through the title, channel_style, and subject_description fields. This data is interpolated directly into prompts for image generation and the agent's final response without sanitization or boundary markers (such as delimiters). An attacker could provide malicious text designed to influence the agent's behavior or the generated image content.
Ingestion points: SKILL.md input fields.
Boundary markers: Absent.
Capability inventory: Includes image generation tools and shell command execution (curl).
Sanitization: None identified in the instructions.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using curl to interact with the api.muapi.ai endpoint as a fallback mechanism. While this is part of the skill's intended functionality to communicate with its own API, it represents a functional capability that involves the shell and network access.

muapi-youtube-thumbnail