seedance-continuation
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill primarily consists of instructional logic for media generation workflows, defining how an agent should handle transitions and continuity between video clips.
- [COMMAND_EXECUTION]: The skill contains a reference to a local script
scripts/extract_last_frame.py. This is presented as a text recommendation for the user to execute manually in their local environment to provide image data to the agent; it does not involve automated, remote, or high-privilege execution by the skill itself. - [INDIRECT_PROMPT_INJECTION]: While the skill ingests external data in the form of uploaded clips or frames, it implements safeguards by requiring the agent to use structured observation records and strictly grounded 'canonical' data, minimizing the risk of instructions embedded in media files influencing the agent's behavior.
- [DATA_EXPOSURE_AND_EXFILTRATION]: There are no indicators of unauthorized network operations or access to sensitive local environment variables or credentials. File paths mentioned (
../../references/) are relative and typical for skill-based project structures.
Audit Metadata