The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides systematic guidance for circumventing content safety filters and copyright protections on target platforms like Jimeng and Seedance. Modules such as seedance-filter and seedance-copyright describe techniques like the 'Chinese Prompt Trick' and the use of character archetypes to avoid automated detection of restricted content or intellectual property.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes untrusted user data from the seedance-interview module to construct prompts. Untrusted user input enters the agent context through the seedance-interview/SKILL.md module. The implementation lacks explicit boundary markers or instructions to ignore embedded commands within user-provided visions. The agent's capability inventory across all scripts is focused on high-density prompt generation and technical brief construction. Content sanitization is performed via the seedance-antislop and seedance-filter modules which refine user input into production-ready output.
[EXTERNAL_DOWNLOADS]: The skill documentation references official ByteDance/Volcengine API endpoints such as ark.cn-beijing.volces.com and provides guidance for integrating with standard video utilities including ffmpeg, exiftool, and Topaz Video AI for post-processing.

seedance-20