The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/setup.sh script attempts to execute sudo apt-get install to install system dependencies such as xvfb and GL libraries. This is a privilege escalation pattern that seeks administrative control over the host environment.
[COMMAND_EXECUTION]: The skill's core workflow involves dynamically writing a JavaScript file (scene.mjs) and executing it using node and xvfb-run. This execution of code generated at runtime based on external input is a significant security risk.
[EXTERNAL_DOWNLOADS]: The skill performs npm install for dependencies like three, pngjs, and sharp during its setup phase. It also manually deploys a prebuilt binary file (webgl-abi127-linux-x64.node) from its assets into the node_modules directory.
[PROMPT_INJECTION]: The skill demonstrates a clear surface for indirect prompt injection through its image generation workflow.
Ingestion points: The prompt input parameter defined in SKILL.md is used directly to plan the scene and write the scene.mjs script.
Boundary markers: None; user input is interpolated into logic that generates executable JavaScript code without delimiters or defensive instructions.
Capability inventory: The skill can execute shell commands (bash, node, xvfb-run) and perform file operations (cp, mkdir, and fs writes in the rendering pipeline).
Sanitization: No sanitization, escaping, or validation is performed on the user-supplied prompt before it is used to author script logic.

3d-image-renderer