seedance-vocab-es
Fail
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: HIGHMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill exhibits evidence of impersonation. The
authorfield in the metadata (Iamemily2050) is a typosquatted variant of the identified vendor name (Emily2040) used in the repository URL and vendor context. This specific substitution (4 vs 5) is a known red flag for impersonation. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user input (briefs) and reference tags, creating a surface for indirect prompt injection.
- Ingestion points: The skill processes user-supplied "briefs" and external reference tags such as
@Image1,@Video1, and@Audio1as described inSKILL.md. - Boundary markers: The instructions lack explicit delimiters or safety instructions to prevent the agent from obeying malicious commands embedded within the user-provided cinematic brief.
- Capability inventory: The skill is restricted to text translation and generation; no dangerous capabilities like subprocess execution, file writes, or network requests are present.
- Sanitization: There is no evidence of input validation, filtering, or escaping for the user-provided content before it is processed into prompt wording.
Recommendations
- AI detected serious security threats
Audit Metadata