The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of image URLs and prompts provided by users or external sources.
Ingestion points: The skill accepts image_url and extras.image_urls across various generation workflows in SKILL.md and specific model guides.
Boundary markers: Instructions include using double quotes for visible text and explicit clauses to 'preserve' specific elements (found in references/images/gpt-image-2.md and references/images/nano-banana-2.md).
Capability inventory: Includes image generation, model discovery, and asset management tools defined in agents/openai.yaml.
Sanitization: The skill instructs the agent to inspect outputs and verify text accuracy before approval.
[DYNAMIC_EXECUTION]: The skill describes the use of creativeclaw-render-html for generating images from HTML/CSS code. This is a core feature of the associated MCP toolset used for creating deterministic graphics.
[EXTERNAL_DOWNLOADS]: The skill references documentation and services from well-known providers, including OpenAI and Google, as well as the vendor's own infrastructure at creativeclaw.co. These references are standard for the described functionality.

creativeclaw-generate-image