The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple internal scripts to perform its auditing tasks, including html-visual-gate.mjs for visual hierarchy checks, button-harmony.mjs for button layout analysis, and qc-regression.py for regression testing. These are executed locally via Node.js and Python 3.
[EXTERNAL_DOWNLOADS]: The skill uses a headless browser (Playwright) to fetch and render content from user-specified URLs for auditing. This involves standard network activity to retrieve web resources for analysis.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from external websites, which represents an attack surface for indirect prompt injection. However, this risk is mitigated by the architecture, which uses a deterministic, non-LLM engine for measurements (like contrast and tap-target) and treats the LLM's assessment as advisory rather than a blocking gate.

qc-uiux