designkit-edit-tools

DesignKit Edit Tools

Overview

General-purpose image editing capability set, belonging to the DesignKit atomic capability layer. Each capability has clear inputs and outputs and simple routing, and can be used independently or reused by upper-layer workflows.

Capability List

Capability	Operation Identifier	Status	Description
Smart Cutout	sod	✅ Available	Cut the image subject out from the background
AI Enhance Clarity	image_restoration	✅ Available	Improve image clarity

Intent Recognition

User Wording	Route To
cutout, remove background, cutout, remove background, transparent background, background removal, matting, cutout	sod
enhance clarity, image quality restoration, image enhancement, super-resolution, improve image quality, repair low-resolution, image restoration	image_restoration

Conversational Follow-up Strategy

Smart Cutout (sod)

Missing Information	Whether to Follow Up	Follow-up Script
No image	Must follow up	"Please provide the image to be cut out (local path or URL)."
Background color not specified	Do not follow up	Process according to the system default plan
Aspect ratio and dimensions not specified	Do not follow up	Return at the original image dimensions by default

Typical Conversation:

User: "Help me cut out this image" Agent: "Please provide the image to be cut out (local path or URL)." User: provides image Agent: "Okay, I'll help you cut out the subject of this image." → Execute

AI Enhance Clarity (image_restoration)

Missing Information	Whether to Follow Up	Follow-up Script
No image	Must follow up	"Please provide the image to be enhanced for clarity (local path or URL)."
Clarity level not specified	Do not follow up	High definition by default

Typical Conversation:

User: "This image is too blurry, help me enhance its clarity" Agent: (the user has already provided the image) "Okay, I'll help you improve the clarity of this image." → Execute

Execution

After the parameters are completed, call the unified executor:

bash __SKILL_DIR__/../../scripts/run_command.sh <action> --input-json '<参数JSON>'

Example:

bash __SKILL_DIR__/../../scripts/run_command.sh sod --input-json '{"image":"https://example.com/photo.jpg"}'

bash __SKILL_DIR__/../../scripts/run_command.sh image_restoration --input-json '{"image":"/Users/me/photo.jpg"}'

__SKILL_DIR__ Replace with the absolute path of the directory where this SKILL.md is located. The script will automatically handle local image uploads.

Result Handling

Parse the JSON output by the script:

• ok: true → From media_urls extract the result image URL, use ![结果图](url) to display to the user
• ok: false → Read error_type and user_hint, show the user actionable guidance

error_type	User-visible prompt
CREDENTIALS_MISSING	According to user_hint guide the user to configure
AUTH_ERROR	According to user_hint guide the user to verify
ORDER_REQUIRED	Go to Meitu Design Studio to obtain Meidou
PARAM_ERROR	By user_hint fill in the parameters
UPLOAD_ERROR	Check the network or try another image
API_ERROR	Try another image

Boundaries

This skill only performs atomic-level image editing operations. The following scenarios should be handed off to the corresponding skill:

Do not do	Hand off
Generate e-commerce image sets, e-commerce image sets, viral-style image sets, and other multi-step image sets	designkit-ecommerce-product-kit