The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the documents it processes. Instructions embedded within the target document could influence the image generation process or the agent's behavior. \n
Ingestion points: The skill reads the full content of the user-specified document via the Read tool. \n
Boundary markers: Document content is interpolated into prompts using simple headers (e.g., '内容：') without explicit instructions for the AI to ignore embedded commands. \n
Capability inventory: The skill possesses the ability to execute shell commands (Bash), write to the file system (Write), and edit files (Edit). \n
Sanitization: No input validation or sanitization is performed on the text extracted from the document before it is sent to the Gemini API.\n- [EXTERNAL_DOWNLOADS]: The skill interfaces with Google's Gemini API for image generation. These requests target official Google AI service endpoints, which are well-known and trusted.\n- [COMMAND_EXECUTION]: The skill executes local Python scripts (scripts/generate_single_image.py) to manage the image generation lifecycle and API interactions. It also allows for installing dependencies via pip.

lovstudio-document-illustrator