claude-code-video-toolkit
Pass
Audited by Gen Agent Trust Hub on May 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Clones a core repository and templates from a GitHub repository (github.com/digitalsamba/claude-code-video-toolkit) and installs dependencies from Node and Python registries.
- [COMMAND_EXECUTION]: Makes extensive use of shell commands to manage the production lifecycle, including environment setup (npm/pip install), asset processing (ffmpeg), and project orchestration (modal deploy, npm run render).
- [REMOTE_CODE_EXECUTION]: Uses cloud GPU providers like Modal to deploy and execute AI models for speech, image, and video generation, which involves executing code on remote infrastructure.
- [CREDENTIALS_UNSAFE]: Instructions guide the user to provide several sensitive API keys (ElevenLabs, Modal, RunPod, and Cloudflare R2) as environment variables to enable cloud features and file storage.
- [PROMPT_INJECTION]: The skill ingests user-provided scripts (script.md) and configurations (project.json) to generate content, creating a surface for indirect prompt injection if the inputs contain malicious instructions.
- Ingestion points: Processes markdown scripts for voiceover and JSON files for scene orchestration.
- Boundary markers: No specific delimiters are mentioned to separate instructions from user-provided content in script parsing.
- Capability inventory: Includes subprocess execution, file system writes for media assets, and network calls to AI APIs.
- Sanitization: No explicit sanitization or validation of script content is documented in the current workflow.
Audit Metadata