The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/video_transitions.py executes system-level ffmpeg commands via subprocess.run. It dynamically constructs complex filter chains (e.g., fade, zoompan, crop) using values loaded from animation_plan.json. While passing the command as a list prevents shell-level command injection, the lack of strict validation on the configuration values could allow for FFmpeg filter injection if the input plan is malicious.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted JSON data from sources like ppt_data.json and animation_plan.json across several scripts.
Ingestion points: scripts/html_animations.py, scripts/video_transitions.py, and scripts/animation_planner.py read and process external JSON files.
Boundary markers: There are no specific delimiters or instructions used to prevent the agent from treating data inside these JSON files as instructions.
Capability inventory: The skill can write files to the local filesystem and execute FFmpeg commands via subprocess.
Sanitization: In scripts/html_animations.py, content such as slide titles and list items are interpolated directly into an HTML template using string replacement. The template also uses dangerouslySetInnerHTML for the slide content. No HTML escaping or sanitization is performed, creating a Cross-Site Scripting (XSS) vulnerability surface in the generated HTML viewer.
[EXTERNAL_DOWNLOADS]: The HTML template assets/templates/enhanced_viewer.html is configured to download React and Framer Motion libraries from the well-known unpkg.com CDN. While these are reputable services, they represent external dependencies that are loaded when the generated viewer is opened.

remotion-video-enhancer