The Agent Skills Directory

[SAFE]: The skill architecture is designed around a self-contained security model. By instructing the agent to inline all CSS, JavaScript, and SVG assets while explicitly banning external links (CDNs, stylesheets, remote images), the skill minimizes the risk of third-party asset injection or user tracking in the generated documents.
[COMMAND_EXECUTION]: The skill includes a Python utility script, scripts/prepare_template.py, which is used to automate the bundling of HTML templates with a shared base CSS. Analysis of the script shows it performs only local file system operations (reading bundled templates and writing output files) and safe string manipulation using regex. It does not utilize any unsafe functions like eval() or os.system() with user-supplied arguments.
[EXTERNAL_DOWNLOADS]: The skill instructions include a mandatory verification step using rg (ripgrep) to check for external link markers (https?://, <link, src=) in the final output. This proactive check ensures that the 'self-contained' rule is enforced before the artifact is delivered to the user.
[PROMPT_INJECTION]: The skill instructions and metadata contain no patterns associated with prompt injection. The instructions focus on document structure and visual style, and do not attempt to override system safety protocols or bypass agent constraints.

generate-visual