The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script runtime/validator_discovery.py uses subprocess.run to execute Python scripts determined at runtime. The target script path can be provided via the UI_DESIGN_SYSTEM_VALIDATOR environment variable or resolved to local repository scripts. The execution is controlled by passing arguments as a list (avoiding shell evaluation) and verifying a specific DESIGN_SYSTEM_VALIDATOR_ROLE marker within the target script's content before execution.
[COMMAND_EXECUTION]: The SKILL.md instructions guide the agent to execute multiple local Python utilities (check_active_instance.py, check_design_freeze.py, scan_design_constraints.py) to manage the design system lifecycle. These scripts perform local file operations like validation against JSON schemas and regex-based scanning of source code for styling consistency.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process a variety of project files, including UI source code (.tsx, .vue, etc.), CSS files, and design configuration YAMLs. This represents an attack surface where malicious content in the project could influence agent behavior. However, the skill mitigates this by using strict JSON schema validation, directory traversal checks (.. path detection), and a safe YAML loader that inherits from yaml.SafeLoader.

ui-template-design