SuperAPI AI Directory
Partner中文AI newsFavorites
Skillswan-cover-plus

wan-cover-plus

General & other

View repository

Updated 10-16 04:58

About this skill

WanCover+

WanCover+ is acontent visual productionClaude Code Skill.

It is suitable for two types of tasks:

  • Image generation: WeChat Official Account cover images, Xiaohongshu cover images, product recommendation images, poster relayout
  • Video generation: text-to-video, static image to smooth dynamic video, reference image / reference video to video

When this skill must be used

When the user's goal ismaking imagesormaking videos, rather than simply writing articles, writing copy, or making summaries, first determine whether it belongs to the following tasks:

  • WeChat Official Account cover image
  • Article header image
  • Xiaohongshu cover
  • product recommendation image
  • poster redesign
  • horizontal to vertical
  • text-to-video
  • image-to-video
  • static image to dynamic video
  • reference image to video
  • reference video to video
  • smooth dynamic video

If the user is only discussing copy and has no clear image or video delivery goal, do not mistakenly trigger this skill.

Important: path resolution

This skill may be installed in different locations. Before executing any command, first determine the skill root directory based on the directory where the currently loaded SKILL.md is located, then replace the <SKILL_DIR> below with the actual path.

1. Image tasks: first determine the nature of the task, then choose the scenario

Step 1: Determine whether it is "generating an image from scratch" or "modifying an existing visual"

  • If the user's focus is "generate an image based on the title/selling point", handle it asgenerating from scratchHandle
  • If the user's focus is "preserve the original subject / preserve the information hierarchy / horizontal to vertical / relayout", handle it asmodifying an existing visualHandle

This judgment takes priority over platform keywords.

Step 2: Then determine the target platform or distribution format

  • Horizontal article header image, WeChat Official Account cover, tech media header image → wechat_cover
  • Vertical feed cover, Xiaohongshu cover, product recommendation image → xiaohongshu_cover
  • Preserve the original visual structure and redesign → relayout_poster

Image scene conflict resolution table

Conflicting input Correct handling
"Xiaohongshu cover" + "preserve the original poster subject" Prioritize relayout_poster, because the essence of the task is redesign, not generating from scratch
"WeChat Official Account header image" + "vertical distribution" First confirm the platform goal; if the final result is a social media vertical card, prioritize xiaohongshu_cover
"make a cover" + "horizontal to vertical" Prioritize relayout_poster
"tech article header image" + "make it more like Xiaohongshu" Prioritize the final distribution platform; for WeChat Official Account publishing prioritize wechat_cover, for Xiaohongshu distribution prioritize xiaohongshu_cover
"Just use this title to generate an image" + "preserve the information hierarchy" If there is no old visual structure, it is still generating from scratch; do not mistakenly select relayout_poster

II. Video tasks: first determine task_type

Video tasks should not mistakenly go through the image scene route; prioritize judging by task_type determine.

text_to_video

Suitable for:

  • The user only provides a text description
  • Want to directly generate a product demo, atmospheric clip, or short video segment
  • Can run even without a reference image or reference video

image_to_video

Suitable for:

  • The user explicitly provides a static image
  • Want to make the static image "move"
  • The focus is preserving the subject, adding slight camera movement and natural motion
  • Default model slot:wan2.7-i2v
  • Also compatible with:wan2.6-i2v-flash
  • Protocol difference: Wan 2.6 uses input.img_url, Wan 2.7 uses input.media, and the image type is first_frame

reference_to_video

Suitable for:

  • The user has a reference image or reference video
  • Hope to preserve subject consistency, reference material style, or motion rhythm
  • Local paths or URLs are acceptable
  • Default model slot:wan2.7-r2v
  • Also compatible with:wan2.6-r2v-flash
  • Protocol difference: Wan 2.6 uses input.reference_urls, Wan 2.7 uses input.media, and the image / video types are respectively reference_image / reference_video

Principles for determining video tasks

  • Only a static image → prioritize image_to_video
  • Has a reference image or reference video, and hopes to preserve reference features → prioritize reference_to_video
  • No reference material at all, only a text idea → text_to_video

III. Minimum input determination

Keep only the minimum determination; do not expand the full field manual in the main file.

Image tasks

At minimum, it must be possible to determine:

  • title
  • scene
  • style

Video tasks

  • text_to_video: at least title or prompt
  • image_to_video: at least reference_images
  • reference_to_video: at least reference_images or reference_videos

If fields are uncertain, do not guess; go read references/input-schema.md。

IV. When to read reference

  • When fields are uncertain: read references/input-schema.md
  • When the image scene is uncertain: read references/scenes.md
  • When API / output / configuration / success determination is uncertain: read references/api-behavior.md

Do not read all references at once.

  • If it is only image generation, do not load the video troubleshooting instructions together
  • If it is only video generation, do not treat scenes.md as video routing documentation
  • When the user has already provided complete JSON, do not reread the field manual just to be safe

V. How to determine whether it truly succeeded

Only when all of the following conditions are met is it truly successful:

  1. The API returns success
  2. The image task obtained decodable or downloadable image content, or the video task obtained video_url
  3. The output file was actually written successfully
  4. .prompt.txt The file was actually generated
  5. In the JSON, output_files points to a file that actually exists

If any item is missing, it should not be described as "successfully generated".

NEVER list

  • NEVER Pretend that images or videos can be generated normally when wan.api_key is missing
  • NEVER Mistakenly trigger this skill when the user has not requested image or video generation
  • NEVER Treat relayout_poster Treat it as the default fallback scenario for all visual tasks
  • NEVER Fabricate brand colors, reference images, reference videos, URLs, or editing instructions when fields are incomplete
  • NEVER Misroute video tasks to the image scene
  • NEVER Forge a "generation successful" result when the API returns a failure
  • NEVER When the .png or .mp4 file has not actually been obtained, claim the task is complete
  • NEVER Treat config.example.yaml as a real configuration and use it directly without reminding the user to add the key
  • NEVER Unilaterally change the user's core theme, reference materials, or task type just to make the result look better

Reference documentation

  • references/README.md
  • references/input-schema.md
  • references/scenes.md
  • references/api-behavior.md