The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes three local scripts (check-contrast.mjs, detect-token-drift.mjs, and validate-metadata.py) designed to be executed by the agent. These scripts are read-only utilities used for calculating color contrast ratios, scanning source code for hardcoded color values (token drift), and validating metadata strings. They do not perform network operations or execute arbitrary remote code.
[INDIRECT_PROMPT_INJECTION]: The skill operates as an auditing tool that ingests user-provided source files (CSS, JavaScript, and TypeScript) through its scanning scripts. This creates a surface for indirect prompt injection if those source files were to contain adversarial instructions. However, the scripts are narrowly scoped to regex-based extraction of color values and do not interpolate data into the agent's execution context in an unsafe manner, making the risk negligible.
[SAFE]: All external references point to official GitHub repositories and domains associated with the author 'pedronauck', which are recognized as vendor resources. The instructions focus on established design heuristics (Nielsen, W3C) and provide standard industry best practices for UI development.

ui-craft