buda-video-studio
Pass
Audited by Gen Agent Trust Hub on Jul 1, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: Authentication is managed through official Google OAuth flows. Credentials and tokens are stored in a dedicated local configuration directory (~/.config/buda-video-studio/), ensuring they remain under the user's control and separate from the skill's code. The skill also includes a release-check script to prevent accidental inclusion of these secrets in distribution packages.\n- [SAFE]: The skill executes local commands for authentication (opening a browser) and OCR (running swift or tesseract). These operations use fixed script paths and validated local file arguments, preventing arbitrary command injection. All external network requests are directed to well-known and trusted Google API endpoints.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it reads folder names and file contents (Markdown/CSV) from Google Drive projects. This finding is assessed as SAFE because: 1. Ingestion points: google-drive-api-reader.mjs and topics-data-source.mjs. 2. Boundary markers: Data is processed for structured display and file generation, not directly as system instructions. 3. Capability inventory: Local file writes and scoped command execution. 4. Sanitization: Filenames are processed via slugify and display content is HTML-escaped.\n- [SAFE]: The project follows the 'App-in-Skill' pattern, providing a clear boundary between the AI agent's instructions and the local execution of tasks. Human review is integrated as a mandatory checkpoint for any approval-gated actions, such as generating production briefs or marking distribution as complete.
Audit Metadata