film-dialogue-voiceover
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes character scripts and dialogue metadata which are subsequently interpolated into external text-to-speech and video generation tools, creating an attack surface for instructions embedded in the script data.
- Ingestion points: Processes approved scripts, character reference metadata, and API responses from media providers in
SKILL.md. - Boundary markers: Relies on manual review status fields and "approved script" designations as logical boundaries.
- Capability inventory: Performs file writing for project manifests (
voice-cast.json,speech-lines.json) and initiates network API calls via configured tools. - Sanitization: Recommends the use of secret-free input records and local asset storage to mitigate data leakage.
- [EXTERNAL_DOWNLOADS]: The instructions reference and recommend the use of official tools and documentation from ElevenLabs and Volcengine to facilitate audio and video generation. These references target well-known technology providers.
Audit Metadata