The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses ffmpeg and ffprobe to process video files. It executes these tools to extract technical metadata and specific image frames for visual analysis. The commands are local and align with the skill's stated purpose of video evaluation.
[PROMPT_INJECTION]: The instructions establish a detailed scoring framework (55-point scale) and reporting format. There are no attempts to override agent safety guidelines, bypass constraints, or extract system prompts.
[DATA_EXFILTRATION]: Analysis of the workflow shows no network-based commands (such as curl or wget) and no patterns suggesting the exfiltration of user data or sensitive configuration files.
[CREDENTIALS_UNSAFE]: The skill does not contain hardcoded API keys, tokens, or passwords, nor does it attempt to access sensitive credential directories like .ssh or .aws.

nl-video-score