The Agent Skills Directory

[SAFE]: The skill provides instructions for using the Magic Hour API to generate product images. It emphasizes accuracy, preservation of product details, and verification against source materials, which are standard best practices.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data, including product photos and listing copy, which are potential vectors for indirect prompt injection.
Ingestion points: Product photos and listing copy are processed in SKILL.md and references/listing-roles.md.
Boundary markers: None explicitly defined in the prompt structure, but the skill instructs the agent to separate verified facts from observations.
Capability inventory: The skill uses ai_image_editor_create_image, ai_image_generator_create_image, and account_retrieve as part of its operations.
Sanitization: The skill mitigates risks by requiring the agent to compare generated images against original references and documentation before finishing the asset. This surface is considered safe given the specific vendor context and task constraints.

magic-hour-marketplace-images