The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the playwright Python package and browser binaries to support its quality assurance workflow. It extensively references external media assets and font libraries from reputable CDNs such as unpkg.com, cdnjs.cloudflare.com, and cdn.jsdelivr.net, as well as assets hosted on the author's domain and CloudFront.
[COMMAND_EXECUTION]: The skill workflow requires the agent to execute a local utility script, scripts/shoot.py, which uses sub-processes to launch a browser. This is an intentional part of the 'screenshot-and-fix' loop designed to verify the visual integrity of generated code.
[DYNAMIC_EXECUTION]: The scripts/shoot.py script executes a hardcoded JavaScript snippet within the browser context to extract page metrics (e.g., console errors, image load status). The script is static and behaviorally focused on verifying the AI's own generated HTML.
[INDIRECT_PROMPT_INJECTION]: The skill processes a user-provided 'brief' to determine design goals. It lacks explicit boundary markers or isolation for this input when interpolating it into prompts. While this creates a theoretical surface for indirect injection, the risk is localized to the generated web content rather than the agent's environment.
[MALICIOUS_URL_EVALUATION]: The scanner detection of michaelsmith.com was found in examples/portfolio-cosmic-hero.md inside a mailto:hello@michaelsmith.com link. This is a common name used as a placeholder in a sample portfolio template and does not represent an active threat in this context.

motionsites-web-builder