video-podcast-maker-lite
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads background music tracks from Pixabay's official content delivery network. As Pixabay is a well-known service for stock media, these downloads are considered standard functionality for a content creation skill.
- [COMMAND_EXECUTION]: The skill executes shell commands for media processing, including
ffmpegfor audio/video mixing andffprobefor duration detection. It also utilizesnpx remotionfor rendering video frames. These commands are essential to the skill's primary purpose and do not involve untrusted code execution. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied topics to generate narration scripts, which creates a potential surface for indirect injection.
- Ingestion points: The agent writes the
podcast.txtscript based on user input (SKILL.md Step 1). - Boundary markers: Narration segments are clearly delimited using
[SECTION:name|label]markers (SKILL.md Step 1). - Capability inventory: The skill possesses the capability to execute shell commands (
ffmpeg,remotion) and interact with the Azure TTS API. - Sanitization: The
scripts/tts.pyscript escapes HTML special characters (&,<,>) before building SSML. Furthermore, the workflow enforces mandatory human review gates after script generation and before final rendering to mitigate automated exploitation. - [SAFE]: No patterns of obfuscation, credential exfiltration, persistence, or privilege escalation were found. Stated functionality matches the implemented code and instructions.
Audit Metadata