The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze existing project code, content models, and rich media to build new components, which constitutes a potential injection surface.
Ingestion points: In SKILL.md, the agent is instructed to inspect target pages, components, design tokens, and content models. references/components/article-content-layout.md specifically covers handling rich HTML and embeds from external sources.
Boundary markers: The instructions include explicit warnings to treat external output as untrusted. For example, references/components/conversation.md states, "Treat rendered model output as untrusted content."
Capability inventory: The skill possesses the capability to modify or create files within the project repository to implement new components.
Sanitization: The guidance emphasizes safety measures such as performing authoritative server-side validation (references/components/task-input.md) and ensuring API credentials remain on the server boundary (references/components/newsletter-signup.md).
[SAFE]: The skill proactively includes defensive security guidance, instructing the agent to avoid embedding secrets or environment-specific paths in the client bundle and to treat rendered HTML and embeds as security boundaries.
[SAFE]: The reference to an external GitHub repository (github.com/kostja94/pagina) belongs to the skill's author and represents a legitimate vendor resource used for page building functionality.

component-builder