Xiaohongshu Cover Maker

On Xiaohongshu (Little Red Book), a compelling cover image directly determines a post's click-through rate. Creators often face specific challenges: an existing photo needs quick beautification to fit the platform's vertical 3:4 aesthetic; a topic idea lacks a concrete visual presentation; or an approved cover requires local corrections to color, lighting, or background. Traditional design workflows are time-consuming and require professional software, hindering rapid iteration.

How the Skill Works

This skill offers an optimized creation path, translating these needs into precise tool calls. Its core capabilities revolve around three technical routes:
- Enhance Photos into Covers: When a user has a source photo (e.g., food, product, outfit, selfie), it uses beatra.images.transform. This processes the photo into a clean, vertical 3:4 Xiaohongshu cover, automatically optimizing lighting, background, and reserving text areas. This is the default and most common route.
- Generate Cover Concepts from Topics: When no source photo exists, it generates high-click cover concepts based on a confirmed post topic (e.g., "Winter Home Essentials") using beatra.images.generate. The skill avoids generating from a single vague style word, instead forming a complete visual brief combining topic, category, target audience, and references.
- Refine Approved Covers: It uses beatra.images.edit for local adjustments on an existing cover, such as fixing color, lighting, or background flaws. It supports up to two normalized edit regions without altering the overall layout.

Key Steps and Considerations

The workflow emphasizes precision and single-execution:
1. Forming a Cover Brief: The skill reuses the user's provided topic, style preferences, title copy, and visual references. It only asks for missing decisions that would materially change the result (e.g., confirming the topic direction if none is provided).
2. Preparing the Call: All operations must be executed via the bundled MCP client (scripts/mcp_client.py). For photo routes, the source photo is uploaded first, and its role is explicitly noted in the prompt. Defaults are model=auto and count=1.
3. Confirmation and Single Execution: Before any paid call, it displays the final prompt, reference images, canvas ratio, style direction, and all parameters. Upon user confirmation, a unique client_request_id is generated for one-time submission. Any parameter change requires a new confirmation and a new call to prevent duplicate charges.
4. Polling and Delivery: After submission, it exclusively uses beatra.tasks.get to poll for results. Upon delivery, it reviews thumbnail clarity, title readability, color/lighting, canvas ratio, and overall appeal, providing the artifact link and billing information.

Applicable Boundaries

This skill is deeply integrated into a specific workflow; note the following:
- Platform Specificity: All visual specifications (e.g., vertical 3:4 canvas, text safe zones) are optimized for Xiaohongshu's aesthetics.
- Tool Dependency: All remote calls must be made using the bundled scripts/mcp_client.py; REST/OpenAPI cannot be used as a fallback.
- Optional Query Feature: Offers paid "Read Xiaohongshu" queries for data like similar posts or comments. Each query requires individual confirmation and billing (e.g., 180 credits for one page of similar posts, one post, and its comments), but the cover can be delivered normally without queries. Data from queries is explicitly sourced.
- Non-General Design Tool: Specialized for generating and refining Xiaongshu covers, not a replacement for general-purpose graphic design software for freeform creation.