The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches essential assets such as fonts, icons, and functional libraries from trusted, well-known services. This includes Google Fonts (fonts.googleapis.com), Lucide icons (unpkg.com), Motion One (cdn.jsdelivr.net), and MapLibre GL for map components. These are standard industry dependencies and do not represent a security risk.
[COMMAND_EXECUTION]: The SKILL.md file contains instructions for the AI agent to manage skill updates using git fetch and git pull. This mechanism is designed to be interactive and requires explicit user confirmation, ensuring transparency in skill maintenance.
[DYNAMIC_EXECUTION]: Diagnostic scripts included in the skill utilize dynamic execution for validation and measurement. validate-presenter-mode.mjs uses the Node.js vm module to parse slide metadata from generated HTML, and validate-swiss-deck.mjs can utilize the playwright library to measure layout integrity in a headless browser. Both are used locally to ensure the quality of the generated output.
[INDIRECT_PROMPT_INJECTION]: The skill generates presentations from user-provided documents, creating a surface for indirect injection if input material contains malicious instructions. The skill mitigates this by using strict HTML layout skeletons, enforcing specific data attributes, and providing validation scripts that check the generated structure against a predefined schema.

guizang-ppt-skill