byted-mediakit-image

image MediaKit Skill

Usage Rules

1. First read ../byted-mediakit-shared/SKILL.md, and perform the unified pre-check; if this Skill is missing, stop and prompt for installation.
2. Only select from the table below image domain tools; similar capabilities are distinguished by each tool's "capability description" and parameter boundaries.
3. Before execution, read the corresponding reference as needed; parameter and result descriptions come from the same reviewed copy, and the complete machine contract is subject to the current CLI --schema .
4. When required parameters, authentication environment variables, or real input resources are missing, ask the user for them; general optional fields may only pass through values explicitly provided by the user, and other optional fields may be accurately determined from clear intent, but must not be fabricated.
5. During execution, set MEDIAKIT_SURFACE=skill, specifying that the invocation source is Skill.
6. During execution, set MEDIAKIT_RUNTIME to the current Agent host, to avoid the CLI being unable to reliably identify the parent Agent.

Image Volume Governance and Cross-Domain Routing

• When the user emphasizes "try not to lose image quality", "quality first", or "reduce volume with high quality", and does not provide an exact volume limit, quality value, or format conversion requirement, first read reference/families/image-size-reduction.md, and prioritize slim-image.
• When the user explicitly provides quality, max_size, output_format, requires format conversion, or explicitly accepts lossy PNG compression, select compress-image according to the same family guide.
• When the user only says "compress" or "make it smaller" and it cannot be determined whether quality is prioritized or exact volume, quality, or format control is required, clarify first.
• If only images are mentioned without stating the business goal, clarify first. Making multiple images into a video, overlaying images on a video, or cropping video frames should be routed to editing; video frame extraction, video understanding, video enhancement, or video subtitle removal should be routed to video.

Tool List

Tool	Description	Supported Modes	Command	Reference
add-image-watermark	Adds image and text visible watermarks to images, suitable for copyright identification and anti-hotlinking in material distribution scenarios.	Cloud	mediakit-cli image add-image-watermark	reference/add-image-watermark.md
adjust-image-color	Adjusts the brightness, contrast, and saturation of the input image, supporting 6 quick adjustment effects: brighten, darken, enhance contrast, reduce contrast, enhance saturation, and reduce saturation. Suitable for scenarios such as basic material optimization, unifying the visual style of content, and creating special atmospheres such as solemn or retro.	Cloud	mediakit-cli image adjust-image-color	reference/adjust-image-color.md
compress-image	Lossy image compression and format conversion tool; select when the user explicitly provides quality, max_size, output_format, requires format conversion, or explicitly accepts lossy PNG compression. If the user emphasizes trying not to lose image quality, quality first, or reducing volume with high quality, slim-image should be prioritized.	Cloud	mediakit-cli image compress-image	reference/compress-image.md
crop-image	Performs multi-mode cropping on the input image, and can execute directional cropping, targeted cropping, custom cropping, or inscribed circle cropping, suitable for multi-device size adaptation, subject retention, product image edge removal, and specified region capture.	Cloud	mediakit-cli image crop-image	reference/crop-image.md
enhance-image	Makes intelligent decisions based on image content understanding to improve the image's resolution, clarity, and color performance.	Cloud	mediakit-cli image enhance-image	reference/enhance-image.md
erase-image	Automatically detect and erase text or common icons in images based on different scenarios. The erased areas are repaired using intelligent inpainting technology, so that the repaired areas blend naturally with the background.	Cloud	mediakit-cli image erase-image	reference/erase-image.md
evaluate-image-quality	Used for image quality assessment, performing subjective and objective quality and aesthetic scoring on input images. Suitable for quality monitoring, low-quality image screening, content moderation, recommendation ranking, and training data cleaning.	Cloud	mediakit-cli image evaluate-image-quality	reference/evaluate-image-quality.md
face-blur-image	Automatically detect all face regions in an image and apply mosaic processing, used to protect face privacy in images with one click. Supports batch face privacy protection scenarios such as social platform content moderation, street view or surveillance footage anonymization, news media material processing, and AI training dataset anonymization.	Cloud	mediakit-cli image face-blur-image	reference/face-blur-image.md
flip-image	Supports horizontal or vertical flipping of a single image.	Cloud	mediakit-cli image flip-image	reference/flip-image.md
gaussian-blur-image	Used for image Gaussian blur; quickly blur images by setting the blur strength. Suitable for scenarios such as weakening private information, creating atmospheric backgrounds, generating preview images, and cover backgrounds.	Cloud	mediakit-cli image gaussian-blur-image	reference/gaussian-blur-image.md
image-ocr	Used for general printed text recognition (OCR), recognizing simplified Chinese and English in images, and providing text block position coordinates and confidence reference.	Cloud	mediakit-cli image image-ocr	reference/image-ocr.md
invert-image	Used for image negative, applying a negative (inverted) effect to the input image, mapping the image's light and dark relationships and colors to the opposite of the original, that is, light-dark inversion and colors converted to complementary colors.	Cloud	mediakit-cli image invert-image	reference/invert-image.md
mosaic-image	Supports mosaic censor on the entire image or a specified rectangular area, with adjustable pixel grid shape and size. Supports covering sensitive content such as faces, ID information, license plates, and chat records.	Cloud	mediakit-cli image mosaic-image	reference/mosaic-image.md
probe-image-metadata	Supports querying four types of image information: metadata, avghue, alpha, and blurhash.	Cloud	mediakit-cli image probe-image-metadata	reference/probe-image-metadata.md
remove-image-background	Automatically recognize and retain the image subject, and generate an image with a transparent background after removing the background, used for image background removal (cutout).	Cloud	mediakit-cli image remove-image-background	reference/remove-image-background.md
resize-image	Used for image resizing, supporting precise resizing by specified width and height, and also resizing by long side, short side, or proportional mode. Suitable for multi-platform material adaptation, cover and thumbnail generation, and batch image preprocessing.	Cloud	mediakit-cli image resize-image	reference/resize-image.md
rotate-image	Rotate images by setting the rotation angle and rotation background style. Suitable for image orientation correction, creative editing, and batch image processing.	Cloud	mediakit-cli image rotate-image	reference/rotate-image.md
round-corner-image	Quickly add circular or elliptical rounded corners to the four corners of an image. Suitable for common visual editing scenarios such as avatars, cards, and e-commerce main images.	Cloud	mediakit-cli image round-corner-image	reference/round-corner-image.md
sharpen-image	Used for image sharpening. By sharpening the input image, it effectively enhances the edge details and overall clarity of the image. Suitable for scenarios such as e-commerce material optimization, UGC quality enhancement, and cover poster secondary creation.	Cloud	mediakit-cli image sharpen-image	reference/sharpen-image.md
slim-image	A quality-first image slimming tool; choose this first when the user emphasizes trying not to lose image quality, quality first, or high-quality size reduction, and does not require an exact size limit, quality value, or format conversion.	Cloud	mediakit-cli image slim-image	reference/slim-image.md
smart-crop-image	Automatically recognize the subject face region in an image and crop it to fit the specified size; supports both normal face and anime face scenarios. When no face is recognized, the result can be output according to a preset fallback strategy.	Cloud	mediakit-cli image smart-crop-image	reference/smart-crop-image.md