The Agent Skills Directory

[SAFE]: The skill instructions and provided templates do not contain any malicious patterns, credential exfiltration, or unauthorized system access. All code snippets and patterns align with the stated purpose of creative typography and web animation.
[EXTERNAL_DOWNLOADS]: The skill fetches creative coding libraries and fonts from well-known and reputable services including esm.sh, jsdelivr.net, and Google Fonts. These are used to load the core pretext library, animation tools like gsap, and UI components like lil-gui.
[DYNAMIC_EXECUTION]: The workflow involves generating self-contained HTML/JavaScript files based on user requirements and verifying them using a local Python HTTP server. This is a standard and safe procedure for creative coding development.
[INDIRECT_PROMPT_INJECTION]: The skill processes user briefs to generate visual demos, which constitutes an attack surface for indirect instructions. However, the risk is minimal as the output is restricted to client-side browser code.
Ingestion points: User prompts describing a desired creative demo (e.g., "text-as-Breakout game").
Boundary markers: Absent; the agent generates code instructions directly from the user's brief.
Capability inventory: File writing (write_file) and local development server invocation (python3 -m http.server).
Sanitization: None; the agent generates script content based on pattern templates.

pretext