The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/package-loader.mjs implements a dynamic package loader that uses the import() function to execute code from Node.js modules downloaded at runtime to temporary directories.
[EXTERNAL_DOWNLOADS]: The skill performs automated downloads of software dependencies from the npm registry using npm install within its utility scripts. It also extensively uses npx to fetch and execute CLI tools for media initialization and rendering.
[COMMAND_EXECUTION]: The skill utilizes system-level command execution through spawnSync in scripts/package-loader.mjs for bootstrapping its own dependencies and re-spawning the execution environment. It also executes various tasks using npx and node in its standard workflow.
[PROMPT_INJECTION]: The skill ingests untrusted external data in the form of audio files and transcripts to generate video overlays, representing an indirect prompt injection surface.
Ingestion points: Audio and transcript data processed in SKILL.md and references/transcript-guide.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded content are provided for the processed media data.
Capability inventory: Presence of spawnSync for system commands and npx for remote tool execution.
Sanitization: The skill does not document explicit validation or sanitization routines for external transcript content.

hyperframes