The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains logic in SKILL.md and agents/seedance-runner.md directed at evading the safety filters of the Seedance 2.0 model. It provides a specific mapping of restricted keywords to permissive alternatives (e.g., 'fanservice' to 'slapstick comedy', 'ecchi' to 'anime style') to ensure content moderation systems do not block the generation requests.
[COMMAND_EXECUTION]: The script scripts/orchestrator.py utilizes subprocess.run to execute other Python scripts within the skill's directory (such as nano_banana.py and seedance.py). While the command construction uses a list format to mitigate shell injection, the skill relies on executing local code driven by user-provided inputs like prompts and project descriptions.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to download generated image and video files from remote URLs returned by the Replicate API using the requests library, as implemented in scripts/replicate_client.py and scripts/seedance.py.

seedance-animator