ascii-video
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data including video, audio, and text via ingestion points like SRT parsing and video sampling in references/inputs.md. While it primarily renders this data as ASCII, the ingestion and processing of user-provided text quotes and lyrics without explicit boundary markers or sanitization in the provided snippets constitutes a surface area for indirect prompt injection.
- Ingestion points: Media files and text inputs are loaded in references/inputs.md.
- Boundary markers: The documentation and code do not show delimiters or instructions to ignore commands within processed text.
- Capability inventory: The skill has the ability to execute shell commands via ffmpeg and perform network requests to the ElevenLabs API.
- Sanitization: No evidence of text sanitization is provided in the code components.
- [COMMAND_EXECUTION]: The skill uses subprocess.run and subprocess.Popen to interact with ffmpeg and ffprobe for media decoding and encoding as described in references/inputs.md and references/optimization.md. This is a core part of its video production functionality.
- [DATA_EXFILTRATION]: The skill includes functionality to send text data to the ElevenLabs API (api.elevenlabs.io) for text-to-speech generation using the requests library. This represents a legitimate integration with a well-known service for its creative narration feature.
Audit Metadata