Viral Video Script Generator for WeChat Channels

What Problem Does It Solve

When promoting products on WeChat Channels (Video Accounts), creators need to produce short video scripts frequently, especially for "hanging cart" videos that embed product links. Manually writing scripts is time-consuming and often fails to capture the patterns of viral content, leading to unstable conversion rates. This skill automates the generation of high-quality scripts, focusing on improving script relevance and appeal to lower the barrier to content creation.

Core Capabilities and Workflow

Authentication and Setup

The skill calls a remote paid API, requiring a valid API Key for authentication. The key is read from the LY_API_KEY field in config.json, falling back to the LY_API_KEY environment variable if not found. If SSL certificate verification errors occur, you can set the LY_SKIP_SSL_VERIFY=1 environment variable to bypass temporarily, but be aware of man-in-the-middle attack risks. On authentication failure, you must re-obtain the key and update the configuration.

Information Collection and Parameter Mapping

Before script generation, necessary parameters are collected using a two-step process for user-friendliness:
1. Fetch Configuration Options: First, call the API to retrieve all enumerable field options, presented to the user as Chinese labels (e.g., direction options like "达人带货" or "品牌账号").
2. User Selection and Mapping: After user selection, map the Chinese labels back to English value for the backend to avoid 400 errors. Key parameters include:
- mode: Creation mode, supporting creation (original) and recreation (based on existing videos).
- direction: Creation direction, such as influencer promotion, IP account, etc.
- industry and purpose: Industry and marketing goals, at least one required.
- platform: Defaults to channels (WeChat Channels).
- product: Required in creation mode when direction is influencer promotion, needing a product info object with name, price, selling points, and target users.
- source_video: Required in recreation mode, must be a share link from WeChat Channels, passed via --share-url.

Task Creation and Polling

The script generate_script.py handles the entire process:
- Assembles the request body, injecting origin and origin_method identifiers, and automatically validates parameter compliance.
- Sends a POST request to create a task, obtaining a script_task_id, and starts polling to monitor status.
- Polls at fixed intervals, checking the data.status field (e.g., completed or failed), with progress output to standard error and latest responses saved to a JSON file in real-time to prevent timeout loss.

Result Delivery and Billing

Upon task completion, the script render_report.py parses the data.result.markdown field from the response to generate a Markdown report. Delivery includes:
- Displaying actual points used, extracted from the ===POINTS_USED=== marker in standard output; if empty, actual deduction is based on account records.
- Providing the report file path (===MD_PATH===) and a content preview (===OVERVIEW_MD===); it's recommended to open the Markdown file directly for full content.
- On failure, points are refunded uniformly, and users are prompted to resume an existing task using the --task-id parameter to avoid duplicate charges; for example, timeouts or task failures are uniformly explained as network issues with point refunds.

Applicability and Caveats

• Recreation Source Limit: Only supports share links from WeChat Channels; local video file uploads are not accepted. Users must provide a valid share_url.
• Paid Model: Each task consumes API points, estimated around 98 points but actual usage varies. User confirmation of estimated points is required before task creation, and actual points are reported after success or failure.
• Parameter Strictness: Enumerable values must use English value (e.g., direction as influencer); the backend rejects Chinese labels. Validation failures prompt with allowed values in the format "Chinese label (English value)".
• Error Handling: Different strategies are applied based on exit codes (e.g., exit code 8 for invalid key requires re-obtaining, exit code 5 for timeout allows resuming). Uniform scripts avoid transmitting backend error details to users.
• Report Output: The final Markdown report is directly from the backend response and should not be manually modified; to adjust content, modify input parameters (like account_brief or additional_requirements) and regenerate the task.