The Agent Skills Directory

[SAFE]: No malicious patterns, prompt injections, or unauthorized data exfiltration attempts were detected. The skill's behavior strictly adheres to its described functionality as a video post-production tool.
[EXTERNAL_DOWNLOADS]: The skill instructs users to install well-known libraries (sharp, puppeteer, gsap) via the official Node Package Manager (NPM). It additionally utilizes uvx to execute whisperx for transcription. These resources originate from trusted ecosystems and are standard for this application domain.
[COMMAND_EXECUTION]: Automated scripts such as matte.cjs, audio-envelope.cjs, and transcribe.cjs make frequent use of child_process to execute local media utilities like ffmpeg and ffprobe. These operations are performed on user-provided local source files to extract metadata, audio, or video frames.
[OBFUSCATION]: The skill employs Base64 encoding specifically to inline font assets (WOFF2 files) into generated CSS and HTML via inject-fonts.cjs. This is a documented, legitimate technique used here to ensure deterministic rendering results across different systems without relying on installed system fonts.
[DYNAMIC_EXECUTION]: The layout measurement system (measure-layout.cjs) uses a headless browser to evaluate the rendered size and position of text elements. While this involve executing JavaScript in a browser context, it is constrained to local HTML files generated by the skill and serves a functional purpose for occlusion detection.

embedded-captions