The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided descriptions and style instructions to construct prompts for the Gemini model via the mcp__gemini__gemini-query tool. This creates a surface for indirect prompt injection where malicious instructions could be embedded in the user's design request.
Ingestion points: User descriptions and design requirements are ingested in the "Build the Prompt" step (SKILL.md).
Boundary markers: The prompt construction logic in SKILL.md lacks explicit boundary markers or directives to ignore embedded instructions for the user-supplied content.
Capability inventory: The skill uses mcp__gemini__gemini-query for external communication and a file write tool for saving generated SVGs.
Sanitization: The workflow incorporates a dual-model loop where Claude analyzes and refines the SVG output, providing a layer of validation.
[EXTERNAL_DOWNLOADS]: The skill requires and recommends the installation of the @rlabs-inc/gemini-mcp MCP server using npx, which downloads and executes code from the NPM registry. Additionally, the skill's installation command in metadata.json uses curl to fetch the SKILL.md file from a GitHub repository.

gemini-svg-creator