ai-media
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text, such as narration scripts and creative goals, and utilizes it to generate API calls and FFmpeg shell commands.
- Ingestion points: User-provided creative goals and narration scripts defined in the media pipeline in
SKILL.md. - Boundary markers: Absent; there are no instructions to use delimiters or ignore embedded instructions when processing user-provided narration.
- Capability inventory: The skill can write files to the local system (e.g.,
vo.mp3), perform network operations via the ElevenLabs SDK, and generate executable shell scripts. - Sanitization: There is no evidence of sanitization or escaping of user input before it is interpolated into shell command templates or Python code snippets.
- [COMMAND_EXECUTION]: The skill is primarily focused on generating and providing the user with executable shell scripts containing complex FFmpeg filter graphs and command pipelines.
- [EXTERNAL_DOWNLOADS]: The skill references the ElevenLabs Python SDK and documents interactions with several well-known third-party AI media generation platforms (e.g., Google Veo, Kling, Runway, Suno).
- [DYNAMIC_EXECUTION]: The skill generates shell scripts and assembly recipes at runtime based on the specific modalities and parameters required for the user's creative project.
Audit Metadata