The Agent Skills Directory

[SAFE]: No security issues detected. The repository follows best practices for isolation and security.
[COMMAND_EXECUTION]: The script scripts/extract_last_frame.py uses subprocess.run to call ffmpeg for extracting video frames. This is a local utility documented for video processing and does not pose a security risk in this context.
[EXTERNAL_DOWNLOADS]: The test harness scripts/eval_run.py makes connections to a well-known service (Anthropic API) to evaluate prompt quality. This is an optional developer tool that requires user-provided credentials.
[PROMPT_INJECTION]: The skill incorporates robust defensive mechanisms against prompt injection, including sub-skills dedicated to filtering unsafe content and protecting intellectual property.

seedance-20