The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its ingestion of user-supplied topics and external SDK-provided CSS environment variables (Play.fun) without explicit sanitization or boundary markers. This data could potentially influence the agent's code generation logic.
Ingestion points: User-provided topic argument in SKILL.md and SDK CSS variables (--ogp-safe-top-inset, --ogp-safe-bottom-inset) in SKILL.md.
Boundary markers: Absent.
Capability inventory: Shell command execution (via npm and vite) and file system access for project scaffolding.
Sanitization: No sanitization logic was found for external environment data.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of project dependencies (three, vite) from the npm registry and references Three.js library files from the JSDelivr CDN. These are well-known and reputable services used for legitimate software development.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for project setup and management, including directory creation (mkdir), package management (npm install), and starting a development server (vite).

threejs-game