md2we-ai-assistant

name	md2we-ai-assistant
version	1.0.0
description	Generate title suggestions, summaries, and cover images through MD2WE AI endpoints. Use when the user asks for AI-assisted article titling, concise Chinese summaries, or a share-cover illustration derived from Markdown content.
allowed-tools	Read, Write, Edit, Bash, Glob, Grep
user-invocable	true
triggers	["/md2we-ai","generate md2we ai title","generate md2we ai summary","generate md2we ai image","用 md2we 生成标题","用 md2we 生成摘要","用 md2we 生成配图","生成公众号标题摘要封面"]

md2we-ai-assistant

Use MD2WE online AI endpoints for title suggestions, summary generation, and image generation.

When To Use

Use this skill when the user wants to:

• Generate five Chinese title options from an article draft
• Generate a concise Chinese summary for share cards or WeChat digest text
• Generate a 16:9 editorial illustration for the article

AI Config

Requests can pass ai_config explicitly, or rely on server-side defaults.

Supported environment variables:

OPENAI_API_KEY=...
OPENAI_BASE_URL=...
OPENAI_TEXT_MODEL=...
OPENAI_IMAGE_TOOL_MODEL=...

Server-side defaults only support OPENAI_API_KEY, OPENAI_BASE_URL, OPENAI_TEXT_MODEL, and OPENAI_IMAGE_TOOL_MODEL.

If text and image generation need different credentials or base URLs, pass them explicitly inside ai_config in the request body. Do not rely on OPENAI_IMAGE_API_KEY or OPENAI_IMAGE_BASE_URL as server-side defaults.

Service Check

export MD2WE_BASE_URL="${MD2WE_BASE_URL:-https://md2we.com}"
curl -sS "${MD2WE_BASE_URL}/api/health"

API Call Pattern

Title suggestions:

curl -sS "${MD2WE_BASE_URL}/api/ai/title-suggestions" \
  -H "Content-Type: application/json" \
  -d @<(jq -Rs '{
    markdown: .,
    focus_prompt: "更像媒体标题"
  }' article.md)

Summary:

curl -sS "${MD2WE_BASE_URL}/api/ai/summary" \
  -H "Content-Type: application/json" \
  -d @<(jq -Rs '{
    markdown: .,
    focus_prompt: "突出结论"
  }' article.md)

Image generation:

curl -sS "${MD2WE_BASE_URL}/api/ai/generate-image" \
  -H "Content-Type: application/json" \
  -d @<(jq -Rs '{
    markdown: .,
    focus_prompt: "科技感、克制、媒体头图"
  }' article.md)

Pass explicit AI config when needed:

curl -sS "${MD2WE_BASE_URL}/api/ai/summary" \
  -H "Content-Type: application/json" \
  -d @<(jq -Rs \
    --arg api_key "$OPENAI_API_KEY" \
    --arg base_url "${OPENAI_BASE_URL:-https://generativelanguage.googleapis.com/v1beta/openai}" \
    --arg text_model "${OPENAI_TEXT_MODEL:-gemini-2.5-flash}" \
    --arg image_model "${OPENAI_IMAGE_TOOL_MODEL:-gemini-2.5-flash-image}" \
    '{
      markdown: .,
      focus_prompt: "突出结论",
      ai_config: {
        text: {
          api_key: $api_key,
          base_url: $base_url,
          model: $text_model
        },
        image: {
          api_key: $api_key,
          base_url: $base_url,
          model: $image_model
        }
      }
    }' article.md)

Result Handling

• title returns suggestions
• summary returns summary
• image returns image_url and revised_prompt

If the server reports missing AI credentials, surface that directly and stop. Do not invent fallback output.