Script Hijacking: Production-Level Prompt Injection
Preventing malicious inputs from altering the video narrative.
Steps
- Sanitize all user-provided scripts for 'Character Override' commands.
- Run a 'Safety Pass' on the final script before audio synthesis.
- Block scripts that ask for 'Unfiltered' or 'Explicit' vocal tones.
- Use a 'Narrative Guardrail' to keep agents within the defined story.
- Log 'High-Risk' script changes to an administrative dashboard.