The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing the zjandrew-seedance-cli package from PyPI via uv or pipx. This package is the primary tool for the skill's functionality and is directly associated with the skill's author.
[COMMAND_EXECUTION]: The skill instructions involve executing seedance-cli for video generation, configuration management, and task polling. It also provides ffmpeg command templates for frame extraction to allow the agent to inspect generated content.
[CREDENTIALS_UNSAFE]: The skill handles ARK_API_KEY configuration. It advises using environment variables or a specific CLI configuration command that masks the key, and it explicitly warns against storing credentials in shell history, aligning with security best practices.
[INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect injection as it processes user-provided prompts and external media assets (images, videos, audio) which are then passed to the seedance-cli subprocess.
Ingestion points: User-supplied text prompts and local/remote media paths (SKILL.md).
Boundary markers: None explicitly defined for user data in the CLI command construction.
Capability inventory: Subprocess execution of seedance-cli and ffmpeg (SKILL.md).
Sanitization: Not explicitly handled in the skill's markdown; relies on the external CLI tool and the underlying API service's content filtering.

seedance