The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill requires the creation and execution of temporary scripts to extract design tokens from the local codebase as described in RULE-01 and W02.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands such as npx vitest and node references/discover.mjs to perform design analysis and component discovery.
[INDIRECT_PROMPT_INJECTION]: The skill ingests content from external URLs, creating a surface for indirect injection attacks where malicious content on a scraped site could influence the agent.
Ingestion points: The discover.mjs script fetches and screenshots content from user-provided URLs.
Boundary markers: None identified in the processing of external HTML/text to prevent execution of embedded instructions.
Capability inventory: File system writes, shell command execution, and Playwright browser control.
Sanitization: Basic HTML cleanup is performed in discover.mjs for data reduction, but it lacks specific security sanitization against malicious prompt payloads.

ui-kit-from-code