The Agent Skills Directory

[PROMPT_INJECTION]: An indirect prompt injection surface was identified where user-controlled input is interpolated into prompts for downstream image and text generation tools.
Ingestion points: User input for topic, goals, target audience, and channel extracted in Step 1 of SKILL.md.
Boundary markers: Absent; templates in references/prompt-templates.md do not use delimiters or instructions to ignore embedded commands within user content.
Capability inventory: The skill invokes multiple external tools including moai-media:media-higgsfield-image, moai-writer:korean-humanize, and moai-marketer:content-copywriting.
Sanitization: No evidence of input filtering or escaping before data is passed to tool prompt templates.
[EXTERNAL_DOWNLOADS]: The skill documentation and design guides reference external CSS and font resources from well-known services.
Evidence: Links to the jsDelivr CDN (cdn.jsdelivr.net) and Google Fonts (fonts.googleapis.com) for loading the 'Pretendard' and 'Noto Sans KR' font families in references/card-news/design-guide.md.
[SAFE]: The skill follows modular design patterns by delegating tasks like image generation and marketing review to other specialized agent skills within the vendor's ecosystem (e.g., moai-media, moai-marketer). This reflects standard operational behavior for complex AI workflows.

content-card-news