The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill generates HTML diagrams based on user descriptions and writes them to the local disk, creating a surface for XSS if input is malicious.
Ingestion points: System architecture descriptions provided by the user in the first step of the workflow.
Boundary markers: No specific delimiters or safety instructions are defined to isolate user input from the HTML structure.
Capability inventory: The skill uses the write_file tool to save the HTML content to the filesystem.
Sanitization: The instructions do not define any escaping or validation logic for the user-supplied content.
[EXTERNAL_DOWNLOADS]: The skill references external assets and repositories.
The HTML template loads typography from Google Fonts (fonts.googleapis.com), which is a well-known service.
The SKILL.md file provides a reference link to a public GitHub repository for documentation.
[COMMAND_EXECUTION]: The skill advises the user to run shell commands (open or xdg-open) to view the output. While these are suggestions for the user rather than automated actions by the agent, they facilitate interaction with the generated content.

architecture-diagram