The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes external video content via URLs, which creates a surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context through the video_urls parameter in the curl examples within SKILL.md.
Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions that might be embedded in the video file's metadata or content.
Capability inventory: The skill documentation indicates the agent performs network operations and handles tool calls initiated by the API response.
Sanitization: There is no evidence of input validation or sanitization for the content of the external video files.
[DATA_EXFILTRATION]: The skill provides examples for using curl to communicate with sense.eachlabs.run. This network activity is necessary for the skill's primary function and targets the author's verified infrastructure.
[CREDENTIALS_UNSAFE]: Authentication is managed through the $EACHLABS_API_KEY environment variable, which is a standard and safe practice for secret management.
[SAFE]: All external resources and API endpoints belong to the skill's author, eachlabs, and represent legitimate vendor functionality.

video-noise-reduction