The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a 'Figma to Tailwind theme workflow' which introduces an indirect prompt injection surface by ingesting untrusted data to generate code and files.
Ingestion points: In rules/figma-to-theme-workflow.md, the agent is instructed to parse CSS custom properties pasted into the chat by users.
Boundary markers: The skill lacks explicit instructions for the agent to use delimiters or 'ignore embedded instructions' warnings when handling user-pasted data.
Capability inventory: The agent is given the capability to write generated CSS content to multiple files within the local project directory (e.g., src/styles/figma-tokens/).
Sanitization: While the workflow includes parsing and mapping rules, it does not specify sanitization or validation steps to prevent the execution of malicious instructions embedded in CSS comments or variable names.
[COMMAND_EXECUTION]: The skill references standard command-line operations for project setup and build optimization.
Evidence: Files rules/core-utility-model.md and rules/perf-purging-and-scanning.md mention the use of npm install tailwindcss, npm run build, and npx tailwindcss. These are well-known development tools appropriate for the skill's purpose.

tailwind