The Agent Skills Directory

[SAFE]: The skill contains no malicious patterns, prompt injections, or obfuscation. The provided instructions and code snippets follow industry best practices for computer vision tasks.
[EXTERNAL_DOWNLOADS]: Includes commands to install official OpenCV packages (opencv-python, opencv-contrib-python, opencv-python-headless) from legitimate package registries.
[COMMAND_EXECUTION]: Provides standard code examples for local file system access (reading and writing images/videos) and hardware device interaction (camera capture) which are core functions of the OpenCV library.
[INDIRECT_PROMPT_INJECTION]: While the skill establishes a surface for processing external image and video data, this is the intended primary purpose of the tool. The instructions include best practices, such as validating that data has been correctly loaded (e.g., checking if img is not None), to mitigate common processing errors.
Ingestion points: Data is read via cv2.imread() and cv2.VideoCapture() in SKILL.md.
Boundary markers: None explicitly defined, as these are code snippets for agent use.
Capability inventory: File system write access via cv2.imwrite() and cv2.VideoWriter(), alongside image processing algorithms.
Sanitization: Recommends validation checks for loaded resources.

opencv