The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including user-supplied prompts, external image URLs, and web search results when enable_web_search is active. This data influences image generation and tool operations.
Ingestion points: The skill utilizes image_url and extras.image_urls for reference images, and extras.enable_web_search to fetch external information (SKILL.md).
Boundary markers: While the instructions recommend using a "Reference manifest" to label images, there are no strict technical delimiters or instructions for the agent to ignore malicious content within those external sources.
Capability inventory: The skill utilizes tools for image generation, file uploading (get_upload_url, upload_asset), and project management (update_film_project, assemble_film).
Sanitization: There are no explicit instructions to sanitize or validate the content of external references or search results before processing.
[EXTERNAL_DOWNLOADS]: The skill configures an MCP tool using a remote transport at https://app.creativeclaw.co/mcp/chatgpt (agents/openai.yaml). This domain belongs to the skill's vendor and is used to provide the underlying tool functionality.

creativeclaw-nano-banana-pro