Media

Warn

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: Multiple scripts within the skill utilize subprocesses to perform media operations. Art/Tools/ComposeThumbnail.ts spawns the magick (ImageMagick) binary to composite images. Art/Tools/Generate.ts executes magick, cwebp, and rm for image optimization and cleanup. Remotion/Tools/Render.ts uses the Bun shell utility $ to execute npx remotion and npx create-video commands.- [COMMAND_EXECUTION]: The skill implements a mandatory notification system that forces the agent to execute curl requests to http://localhost:8888/notify at the beginning of most workflows. This patterns assumes a specific local service is present and allows for arbitrary message transmission to a local endpoint.- [REMOTE_CODE_EXECUTION]: The Remotion/Tools/Render.ts script includes functionality to create new projects using npx create-video@latest. This command downloads and executes a remote script from the npm registry at runtime.- [REMOTE_CODE_EXECUTION]: The ContentToAnimation.md workflow describes a process where the agent extracts content from untrusted sources (such as YouTube transcripts, PDFs, and web articles via the Parser skill), generates React components containing this data, and subsequently executes npm install and npx remotion render. This represents a dynamic code generation pipeline where external input influences generated source code that is then executed in the local environment.- [EXTERNAL_DOWNLOADS]: The skill interacts with several well-known third-party services for its core functionality, including the Replicate API (for Flux and Nano Banana models), OpenAI API (for GPT-image-1), Google GenAI API (for Gemini/Nano Banana Pro), remove.bg API (for background removal), and the Discord API (for Midjourney integration via a bot).- [CREDENTIALS_UNSAFE]: The image generation scripts (Generate.ts and GenerateMidjourneyImage.ts) are designed to read sensitive API keys from a local environment file (.env) located in user-controlled configuration directories (~/.claude/ or ~/.config/PAI/). While intended for local infrastructure, this involves regular access to a centralized secret store.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 28, 2026, 11:02 PM