The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The image-prompter agent is configured with the WebFetch tool, allowing it to ingest content from external websites to inform the prompt generation process. This creates a surface for indirect prompt injection, where an attacker could place malicious instructions on a website that the agent might fetch to influence its behavior or output.
Ingestion points: The agent uses the WebFetch tool to gather requirements and context as described in agents/image-prompter.md.
Boundary markers: The instructions do not define boundary markers or explicit directives for the agent to ignore instructions embedded within the fetched content.
Capability inventory: The skill documentation includes examples for local file system operations (fs.writeFileSync), API interaction with Gemini models, and the use of image processing CLI tools such as magick and rembg.
Sanitization: There is no evidence of sanitizing or filtering external content fetched via WebFetch before it is processed by the agent.
[EXTERNAL_DOWNLOADS]: The documentation and integration guides recommend the installation of several third-party libraries and CLI utilities for image generation and post-processing.
Evidence: references/integration.md and SKILL.md reference pnpm add @google/genai, pip install rembg, and @imgly/background-removal-node.
Note: These resources originate from well-known services and trusted organizations, and their use is consistent with the skill's stated purpose of image generation and manipulation.

image-gen