SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillspet-commerce-creator

pet-commerce-creator

Coding & development

View repository

Updated 10-15 19:56

About this skill

Cute Pet Product-Selling Content Intelligent Creation Workshop

Task Objective

  • This skill is used for: full-process automated creation of cute pet product-selling short videos, focusing on gaining followers and product conversion
  • Capabilities include:
    • Viral video link parsing and reverse engineering with automatic storage
    • Creative breakdown, script creation, and storyboard design based on the knowledge base
    • Generation of storyboard images, subtitles, and sound effect materials adapted to the COZE API
    • Support for interactive evaluation at every step and batch material download
    • COZE API integrates all materials to synthesize a finished product-selling video
    • Full-process quality inspection and data iteration loop
  • Trigger conditions: user needs for cute pet short video creation (story/review/science popularization/product selling), video link parsing needs, and knowledge base retrieval needs

Prerequisites

  • Dependency description: dependency packages required by the skill 
    coze-workload-identity>=1.0.0  # COZE视频API调用
yt-dlp>=2024.1.0                # 视频下载（支持抖音/B站/YouTube等）
  • Credential configuration:
    • COZE video API credential configured: coze_video_api
    • Credential Key format: COZE_COZE_VIDEO_API_7598017404148613135
    • Usage: automatically read through environment variables, no manual entry required
  • Video download directory:./downloads(automatically created)

Operation Steps

  • Standard process:

    1. The user inputs a video link (Douyin/Bilibili/YouTube, etc.)
    2. Callscripts/video_downloader.pydownload the video and metadata
    3. Callscripts/coze_video_api.pyto handle material upload, video synthesis, and status polling
    4. Execute multi-agent collaborative creation based on the knowledge base reference document (references/agents-prompts.md)
    5. Each agent generates independent outputs with like/dislike/share functions
    6. The finished video supports high-definition download, secondary creation, and platform publishing

  • Video download steps:

    1. The user inputs a video link (supports mainstream platforms such as Douyin/Bilibili/YouTube/Kuaishou/Xiaohongshu, etc.)
    2. Callscripts/video_downloader.py download <URL>download the video
    3. Automatically extract video metadata (title, duration, author, view count, like count, etc.)
    4. Automatically download the video thumbnail
    5. Output the video file path and complete metadata JSON
    6. The Viral Analysis Poop Scooper performs parsing and reverse engineering based on the downloaded video

  • Core agent collaboration:

    1. Viral Analysis Poop Scooper - video link parsing and reverse engineering, knowledge base operations
    2. Cute Fun Creativity Poop Scooper - creative breakdown, cute point extraction, product placement point design
    3. Story Script Meow Manager - product-selling-oriented script creation, shot breakdown
    4. Storyboard Script Shadow Painter - COZE API-adapted standardized storyboard script
    5. Storyboard Director Meow Sir - visual specification design, product display composition planning
    6. Cute Lens Painter Furball Artisan - storyboard image generation, batch download support
    7. Subtitle Cute Editor Pet Sauce - subtitle parameter package, product-selling voiceover text script creation
    8. Sound Effect Fun Matcher Woof Sauce - sound effect material package, voiceover material package creation
    9. Video Composition Director - COZE API full-material integration, sales-oriented finished video composition
    10. Quality Control Inspection Pet Butler - Full-process quality inspection, API adaptation verification, sales-traffic-oriented verification
    11. Data Iteration Poop-Scooping Officer - Data statistics, iterative optimization suggestions, skill upgrades

  • Mandatory rules for pure API composition:

    1. It is prohibited to use any local dependency packages (moviepy/opencv-python/mediapipe/pillow/numpy)
    2. All materials (images/subtitles/sound effects) must comply with COZE API format requirements
    3. The script passesscripts/coze_video_api.pycalls the COZE API, using requests from coze-workload_identity
    4. If local package invocation behavior occurs, immediately terminate the process

  • Optional branches:

    • When a video link is provided: execute viral hit analysis and reverse engineering → automatic storage → generate reference package
    • When secondary creation is needed: trace back from the finished video to any creation step for modification
    • When knowledge base retrieval is needed: match viral hit references according to sales category/follower growth goals

Resource Index

Notes

  • Strictly follow the pure API composition principle; all technical operations are completed through the COZE API
  • Sales traffic is the core orientation; all creation stages must clearly define traffic hooks, product placement points, and follow guidance points
  • Knowledge base linkage takes effect in real time; viral hit analysis results are automatically stored, and creation-type agents automatically retrieve references
  • Interaction data from all steps is synchronized in real time to the creation optimization data pool as the basis for iterative optimization
  • Materials support single-item/single-file + full-package batch download, with unified and reusable formats
  • Finished videos support high-definition download, secondary creation, and platform publishing, with sales information clearly conveyed

Usage Examples

Example 0: Video Download and Analysis Reverse Engineering (New)

  • Function description: download videos from platforms such as Douyin/Bilibili/YouTube, extract metadata, and perform analysis reverse engineering
  • Execution method: video_downloader.py + Viral Hit Analysis Poop-Scooping Officer
  • Key parameters:
    • Video link: the user provides the video URL
    • Output directory: ./downloads (automatically created)
  • Example process:
    1. The user inputs a video link (such ashttps://www.bilibili.com/video/BV1xx411c7mD）
    2. call video_downloader.py to download the video 
      python scripts/video_downloader.py download "https://www.bilibili.com/video/BV1xx411c7mD"
    3. automatically download the video file (MP4 format)
    4. automatically extract video metadata (title, duration, author, view count, like count, etc.)
    5. automatically download the video thumbnail
    6. Output results include:
      • Video file path:./downloads/视频标题.mp4
      • Metadata JSON:./downloads/视频标题.info.json
      • Thumbnail:./downloads/视频标题.jpg
    7. Viral Analysis Poop Scooper analyzes and reverse-engineers based on the downloaded video
    8. Generate an analysis report and automatically store it in the knowledge base

Example 1: Cute Pet Product Review Sales Video Creation

  • Function Description: Create a cat food review sales video, 15-second vertical screen 9:16
  • Execution Method: Multi-agent collaboration + COZE API synthesis
  • Key Parameters:
    • Creation Type: Product Review
    • Sales Category: Staple Food (Cat Food)
    • Follower Growth Goal: Targeted Pet Owners
    • Duration: 15 seconds
    • Style: Cute Pet Light Comedy
  • Example Process:
    1. Viral Analysis Poop Scooper retrieves the knowledge base and outputs a review sales reference package
    2. Cute Creative Poop Scooper breaks down the creative concept (first 3 seconds show cat food packaging → cat craving → eating footage → follow guidance)
    3. Story Script Meow Manager creates the sales script (voiceover: "This cat food formula is super clean, look how much the cat loves it! Remember to like and follow")
    4. Storyboard Director Meow Sir plans the product display composition (cat food close-up + cat eating action)
    5. Cute Shot Artist Furball Artisan generates 4-6 storyboard images, supports batch download
    6. Subtitle Cute Editor Pet Sauce creates the subtitle parameter package (selling point subtitles: clean formula/no additives, follow guidance: follow to get coupons)
    7. Sound Effect Fun Mixer Woof Sauce creates the sound effect + voiceover asset package (light background music + clear voiceover tone)
    8. Video Synthesis Camera Officer calls the COZE API to integrate all assets
    9. Quality Control Inspector Pet Butler checks API compatibility and the practicality of the sales traffic guidance
    10. Output the finished video + asset integration ledger, supports download and secondary creation

Example 2: Viral Video Link Analysis and Reverse Engineering

  • Function Description: Analyze a viral Douyin cute pet sales video, reverse-engineer the creation logic, and automatically store it in the knowledge base
  • Execution Method: Viral Analysis Poop Scooper + Automatic Knowledge Base Archiving
  • Key Parameters:
    • Video Link: User provides a link from platforms such as Douyin/Kuaishou/Xiaohongshu
    • Analysis Dimensions: Traffic logic, sales conversion logic, content creation reverse engineering
  • Example Process:
    1. User inputs the video link
    2. Viral Analysis Poop Scooper analyzes the video (duration 100k+, likes 50k+, conversion rate 2.3%)
    3. Generate analysis report:
      • Traffic Hook: In the first 3 seconds, the cat emerges from the cat food bag
      • Sales Placement: At the 5th second, the voiceover mentions the cat food brand
      • Follow Guidance: At the 12th second, the subtitle displays "Follow for perks"
    4. Automatically classify and store by "Pet Breed/Creation Type/Sales Category"
    5. Output a reference package for subsequent creation calls

Example 3: Full-Step Interactive Optimization and Secondary Creation

  • Function Description: Based on the interaction data of the finished video, optimize creation and perform secondary creation
  • Execution Method: Interaction data collection + retrospective modification + re-synthesis
  • Key Parameters:
    • Finished Video: The already generated sales video
    • Interaction Data: Like rate 85%, dislike rate 5%, comment feedback "Product selling points are not prominent"
  • Example Process:
    1. User views the finished video, likes/dislikes/shares/comments
    2. Data iteration: the poop-scooper analyzes interactive data (dislikes are concentrated on "unclear selling points")
    3. Generate optimization suggestions (strengthen product selling point subtitles, add close-up shots of selling points)
    4. User clicks "one-click secondary creation"
    5. Go back to the storyboard script artist Huiyingjiang / storyboard director and cinematographer Meow Sir to modify the product display design
    6. Regenerate materials and synthesize the finished video
    7. Output the optimized finished video and compare the product promotion and traffic-driving effects before and after optimization