The Agent Skills Directory

[COMMAND_EXECUTION]: The utility script scripts/shot.js uses child_process.execSync to launch a headless browser (Chrome or Edge) for taking screenshots. The command string is constructed using string concatenation with arguments directly from process.argv. Specifically, the variables file (input HTML) and out (output PNG) are not sanitized before being passed to the shell. A maliciously crafted filename provided by a user could lead to arbitrary command execution on the host system when the agent runs the quality control step.
[DYNAMIC_EXECUTION]: The scripts/shot.js script performs dynamic code injection. It reads the source code of an HTML storyboard, injects a custom <script> block designed to定格 (freeze) animations and force a re-render, and then writes this to a temporary file for execution. While functional for its purpose, this runtime modification and execution of code increases the attack surface.
[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves processing untrusted user input (storyboard descriptions and scripts) to generate executable HTML files. Because the agent is also instructed to run a shell-based utility (shot.js) on these generated files, a user could provide input that results in a filename containing shell metacharacters, potentially exploiting the command execution vulnerability mentioned above.
[EXTERNAL_DOWNLOADS]: The HTML templates and examples reference external resources from Google Fonts (fonts.googleapis.com). These references are consistent with the skill's purpose of providing cinematic visual styles and are from a well-known service.

video-shot-demos