The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external design data, which creates a surface for indirect instructions to be passed to the agent.
Ingestion points: Data is read from .claude/design/design-system.md and user-provided HTML/CSS exports from the Stitch web interface.
Boundary markers: The skill does not provide specific instructions to the agent to use delimiters or protective boundary prompts when handling these external data sources.
Capability inventory: The agent has the capability to generate and modify screens, interact with design system tokens, and download visual assets (PNGs).
Sanitization: There is no evidence of sanitization or validation of the ingested design system content or HTML code before it is interpolated into agent prompts.
[EXTERNAL_DOWNLOADS]: The skill connects to Google's official Stitch MCP server at https://stitch.googleapis.com/mcp and references the google-labs-code/stitch-skills repository for advanced workflows.

stitch-design