ppt-to-video-workflow
Pass
Audited by Gen Agent Trust Hub on Apr 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute multiple Python scripts (render_slides.py, generate_narration.py, generate_subtitles.py, create_video.py) and the system-level FFmpeg tool for media processing.
- [EXTERNAL_DOWNLOADS]: The skill depends on several third-party Python packages (pillow, pydub, gradio-client, python-pptx) and interacts with an external 'Chatterbox TTS' API for audio generation.
- [DATA_EXFILTRATION]: Narration text extracted from user-supplied slides is transmitted to an external TTS API endpoint via the gradio-client library.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes untrusted text from external files.
- Ingestion points: The narration and title fields within 'slides.json' and the visual content of 'presentation.pptx'.
- Boundary markers: None. The skill does not implement delimiters or safety instructions to prevent the agent from obeying instructions embedded in the slide data.
- Capability inventory: The skill combines shell command execution (via scripts and FFmpeg), network communication (via the Gradio TTS API), and file system read/write access.
- Sanitization: None. Data from the processed files is used directly for rendering images and synthesizing speech without validation or escaping.
Audit Metadata