synthesia
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were identified.
- [EXTERNAL_DOWNLOADS]: The skill references official documentation and services from Synthesia (synthesia.io) for video generation.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external content and brand profiles to generate scripts for the Synthesia API, incorporating robust safeguards.
- Ingestion points: Content skills, brand-profile, and voice-builder metadata in SKILL.md.
- Boundary markers: Mandatory human-in-the-loop approval policy and requirement to lock scripts before generation as specified in SKILL.md and references/scope-and-connections.md.
- Capability inventory: Synthesia API interaction for video production referenced in references/synthesia-2026-reality.md.
- Sanitization: Requirement for native-speaker QA passes and manual review of all outputs defined in references/fit-and-templates.md.
Audit Metadata