The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits a potential surface for indirect prompt injection by design, as it ingest data from external project files to define branding constraints.
Ingestion points: The skill reads 'Design system handoff directories', 'DESIGN.md', token files (CSS/DTCG), and brand voice documents to extract visual identity markers (Step 1 in SKILL.md).
Boundary markers: There are no explicit delimiters or specific 'ignore instructions' warnings mentioned to encapsulate the content read from these external files when they are incorporated into the model prompt.
Capability inventory: The skill has the capability to perform network-based image generation by delegating tasks to the 'moai-media' component.
Sanitization: The skill does not describe specific sanitization or filtering logic for the ingested brand context before it is converted into prompt constraints.
[SAFE]: The skill uses well-defined delegation for its core execution and references reputable design resources.
The skill delegates image generation, cost calculation, and model selection to an external component ('moai-media:media-higgsfield-core'), following a modular and controlled execution pattern.
External references are provided for documentation purposes and point to well-known design platforms (Recraft, Vistaprint) and project repositories.

design-logo