The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's installation instructions require cloning a repository from an unverified GitHub organization (github.com/xiaopu-ai/web-design). This repository contains the primary logic and scripts for the skill, representing a supply chain risk as the source code is not from a trusted vendor.
[COMMAND_EXECUTION]: The skill documentation encourages the execution of shell commands to install dependencies (pip install playwright beautifulsoup4) and run utility scripts. Specifically, the crawl.py script executes automated browser sessions via Playwright to interact with external networks.
[PROMPT_INJECTION]: The skill processes arbitrary external websites to extract 'design tokens', creating a surface for indirect prompt injection. Content from an attacker-controlled website could be used to manipulate the agent's internal logic or influence the generated code output.
Ingestion points: External URLs are processed by scripts/crawl.py and ingested into the design phase.
Boundary markers: There are no documented delimiters or instructions provided to the agent to distinguish between design data and potential malicious instructions embedded in the crawled HTML/CSS.
Capability inventory: The skill has the capability to perform network requests (crawling) and write multiple files to the local file system (HTML, CSS, JS, and Markdown specs).
Sanitization: The analysis found no evidence of sanitization, filtering, or validation of the content retrieved from external URLs before it is processed by the model.

xiaopu-web-design-skill