The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes p5.js and its ecosystem of libraries (p5.sound, CCapture.js, p5.brush, p5.grain) via script tags pointing to well-known and trusted content delivery networks including cdnjs (Cloudflare), unpkg, and jsDelivr.
[COMMAND_EXECUTION]: The skill includes several utility shell scripts for environment management. scripts/setup.sh verifies system dependencies, scripts/serve.sh starts a local HTTP server using either Python's built-in module or Node.js to load local assets, and scripts/render.sh coordinates the headless browser export pipeline.
[PROMPT_INJECTION]: As the skill involves the agent generating executable code (HTML/JavaScript) based on user-provided prompts, it naturally possesses an indirect prompt injection surface. The automated rendering script scripts/export-frames.js uses Puppeteer with flags that disable certain security features, such as --disable-web-security and --allow-file-access-from-files, which are standard requirements for capturing local canvas elements but could be leveraged by maliciously crafted code.
Ingestion points: User-provided creative prompts are used to generate the p5.js sketches in SKILL.md.
Boundary markers: The instructions do not define specific delimiters to wrap user-influenced strings within the generated code.
Capability inventory: Generated sketches run in the browser with access to p5.js APIs and can be rendered headlessly using Puppeteer.
Sanitization: There is no specified logic for sanitizing or escaping user-provided text before it is interpolated into the generated JavaScript code.

p5js