The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/cast.js is vulnerable to command injection. It builds shell commands for child_process.execSync using string interpolation with the outputPath variable. Since this variable is derived directly from the --output command-line argument without sanitization, an attacker or malicious content could provide a path containing shell metacharacters (e.g., double quotes and semicolons) to execute arbitrary commands in the environment.
[PROMPT_INJECTION]: The skill implements a pipeline that ingests untrusted data from URLs or pasted text and renders it into a browser environment via Playwright. This architecture presents an indirect prompt injection risk; malicious instructions embedded in a processed webpage or text could attempt to manipulate the rendering logic, influence the agent's generation of HTML and CSS, or exploit the headless browser context.
[EXTERNAL_DOWNLOADS]: The project project uses the playwright library as its primary rendering dependency. Playwright is a standard industry tool that fetches browser binaries (such as Chromium) from official Microsoft distribution servers during its initialization process.

cast