seedance-interview
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: Minor discrepancies observed in metadata fields. The author name 'Iamemily2050' differs slightly from the repository owner 'Emily2040' in the provided GitHub URL. Additionally, the 'updated' date is set to a future value (2026-05-30), which may indicate a placeholder or incorrect versioning tracking.
- [INDIRECT_PROMPT_INJECTION]: The skill features a data ingestion surface where it processes untrusted user input describing video ideas to generate creative briefs.
- Ingestion points: User-provided video concepts and ideas mentioned in the 'Process' section.
- Boundary markers: None explicitly defined for user input interpolation.
- Capability inventory: No dangerous capabilities (file writes, network operations, or subprocess execution) are requested or defined in this skill file.
- Sanitization: No explicit sanitization or validation of user input is mentioned in the instructions.
Audit Metadata