The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to perform visual verification by running the shell command npx playwright screenshot. This allows the agent to execute shell processes in the workspace environment.
[EXTERNAL_DOWNLOADS]: The use of npx facilitates the download and execution of the Playwright package and its browser dependencies from the official npm registry if they are not already present.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it processes untrusted data from external sources while having the capability to execute shell commands.
Ingestion points: The skill retrieves design context from Figma URLs via MCP tools (get_design_context, get_screenshot) and reads local project files such as .agents/design-context.md and tailwind.config.*.
Boundary markers: There are no specific instructions or delimiters defined to prevent the agent from interpreting text found within Figma files or context files as high-priority instructions.
Capability inventory: The skill provides the agent with shell execution capabilities through the Playwright verification step.
Sanitization: The instructions do not include any input validation or sanitization steps for data fetched from the Figma API or local configuration files before it is processed by the agent.

ui-design