The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes external project data such as CSS, configuration files, and component source code to generate a style guide and update agent instructions. This workflow creates an attack surface where malicious instructions embedded in the project source code could be promoted into the agent's persistent instruction files.
Ingestion points: Reads CSS custom properties, Tailwind/theme configurations, and component library source code (SKILL.md, Step 2).
Boundary markers: The instructions do not direct the agent to use boundary markers or delimiters to isolate extracted content from the system instructions in the resulting files.
Capability inventory: The skill allows the agent to write DESIGN.md and modify project-specific configuration/instruction files like AGENTS.md, CLAUDE.md, and .cursor/rules (SKILL.md, Steps 3 and 4).
Sanitization: There are no instructions for sanitizing, escaping, or validating the text extracted from the project files before it is used to generate new instructions.
[EXTERNAL_DOWNLOADS]: The skill documentation guides the user to connect an optional external MCP (Model Context Protocol) server.
Evidence: Mentions the configuration and use of an MCP server at https://uxkin.com/mcp for searching design systems.

design-md