seedance-2-0
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied data through the
PROMPTvariable within its operational examples inSKILL.md. While the skill provides templates for structured prompting, it lacks explicit boundary markers or instructions to treat user input as untrusted data. - Ingestion points: User input is ingested via the
PROMPTvariable in thevideo_selectorandseedance_videotool execution examples. - Boundary markers: Absent. There are no delimiters or instructions provided to isolate or ignore potentially malicious instructions within the user-supplied prompt.
- Capability inventory: The skill environment is configured with
allowed-tools: Bash, Read, Write, providing a significant surface for exploitation if a user-supplied prompt contains successful injection attacks. - Sanitization: Absent. The provided logic does not include sanitization, validation, or escaping of the user-supplied content before it is processed by the agent tools.
Audit Metadata