dynamic-promo-video
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to research external software projects (GitHub repositories, package registries) and use that information to generate narration and video scenes. This ingestion of untrusted external content creates a vulnerability surface where malicious instructions hidden in a project's documentation could attempt to influence the agent's behavior.
- Ingestion points: Researching project READMEs and documentation (SKILL.md, Step 0; references/screenshots.md).
- Boundary markers: The instructions lack explicit delimiting or 'ignore' warnings for the researched content, though they advise the agent to use 'real names' and 'do not invent features'.
- Capability inventory: The skill uses file-writing, network communication (TTS API), and system command execution (ffmpeg, node).
- Sanitization: No specific sanitization of external text is performed before it is used in narration or HTML templates.
- [COMMAND_EXECUTION]: Several scripts (
mix.py,verify.py,timing.py,demo.py) usesubprocess.runto call system tools likeffmpegandffprobefor media processing and verification. These calls use the safe list-based format for arguments rather than shell strings, which significantly reduces command injection risks. - [EXTERNAL_DOWNLOADS]: The script
tts_elevenlabs.pycommunicates with the ElevenLabs API (api.elevenlabs.io) via HTTPS POST to generate narration audio. ElevenLabs is a well-known service, and the use is documented and relevant to the skill's primary function of video creation.
Audit Metadata