The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a workflow where sub-agents execute shell commands including npm run build, npm run lint, and npx tsc. These commands run scripts defined in the user's project environment, which could execute arbitrary code if the project configuration is compromised.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources such as Figma URLs, remote design URLs, and local image files to drive UI implementation.
Ingestion points: External content from design references (design_reference) and live implementation previews (app_url) are loaded into the agent's context.
Boundary markers: The instructions lack explicit delimiters or safety instructions that direct the sub-agents to ignore potentially malicious text embedded within the visual or metadata components of design references.
Capability inventory: The skill has access to Bash, Read, and file-writing capabilities through its sub-agents, creating a path for misuse if a malicious design reference successfully influences the agent's output.
Sanitization: No specific sanitization or filtering logic is mentioned for content extracted from Figma or remote websites before it is provided to the code implementation sub-agents.

ui-implementer