The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to read the contents of the toolkit's .env file using cat and grep to verify configuration. This operation accesses a sensitive file path that may contain credentials or other private environment variables.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data to generate video assets, creating an injection surface.
Ingestion points: Video briefs, projects/PROJECT_NAME/src/config/demo-config.ts, and projects/PROJECT_NAME/VOICEOVER-SCRIPT.md.
Boundary markers: No boundary markers or 'ignore' instructions are provided to delineate untrusted input.
Capability inventory: The toolkit includes Python script execution, npm install capability, and media processing via ffprobe.
Sanitization: No explicit sanitization or validation of the input content is described.
[EXTERNAL_DOWNLOADS]: The skill installs dependencies from NPM and PyPI and interacts with the Modal cloud platform for GPU-accelerated tasks.
[COMMAND_EXECUTION]: The toolkit makes extensive use of shell commands. It specifically uses the --break-system-packages flag with pip3, which can modify the system's Python environment and potentially cause conflicts with system-managed packages.

video-toolkit