The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied references and instructions to influence the generation of p5.js code. While this creates an attack surface where instructions could be embedded in user input, the skill provides a structured framework requiring the creation of an "algorithmic philosophy" first, which acts as a conceptual boundary before code implementation. Mandatory evidence chain: 1. Ingestion points: User instructions provided as a reference (SKILL.md). 2. Boundary markers: The agent is instructed to first generate a markdown philosophy and then use that to guide implementation. 3. Capability inventory: Generates p5.js code inside a self-contained HTML artifact with UI controls. 4. Sanitization: The skill relies on template-based structure but does not specify input sanitization.
[EXTERNAL_DOWNLOADS]: The interactive viewer template fetches the p5.js library from Cloudflare's CDNJS and typography from Google Fonts. These are well-known and trusted services used to provide core functionality and consistent branding for the generative art viewing experience.
[SAFE]: Analysis of the skill's instructions and templates revealed no evidence of data exfiltration, command execution, obfuscation, or persistence mechanisms. The code generation patterns are standard for interactive art skills on the platform.

algorithmic-art