Social Media Operation Research Report (Paid Version)

Why Automate Social Media Research?

Manually mining Douyin or WeChat Video Accounts for competitor content and analyzing their strategies often involves watching countless videos, taking notes, and compiling reports—a process that can take hours or days. This is inefficient for teams needing quick insights into market trends, campaign performance, or content planning. A solution that automates batch collection and analysis is essential.

How the Skill Works

The skill leverages server-side automation to separate raw material collection from report generation, with a user alignment step in between to refine the analysis direction.

Core Capabilities & Key Steps:

1. Setup & Authentication: Execution requires a valid LY_API_KEY, used to call Zero One Data's APIs for collection and generation. The key is stored in the skill's config.json or injected via environment variables.
2. Parameter Collection & Task Creation: The user provides optional brand/selling points, collection sources (e.g., platform, accounts, keywords, items per source), and an initial research objective. The skill uses create_task.py to create a collection task on the server, estimates the point cost, and proceeds only after user confirmation.
3. Asynchronous Collection & Polling: After task creation, the server asynchronously collects videos and reads their raw content. The skill periodically queries the task status using poll_task.py (e.g., capturing, analyzing) and provides real-time progress updates to the user.
4. Analysis Alignment & Report Generation: Once collection completes (status awaiting_prompt), the skill discusses and refines the specific analysis focus (e.g., competitor benchmarking, script analysis, or topic opportunities). After confirmation, it submits the final research prompt via resume_task.py. The server then generates a structured Markdown report covering executive summary, data foundation, and in-depth analysis.
5. Report Delivery & Validation: The generated Markdown content is retrieved from the server and saved as a .md file. The skill reads back the file to validate its integrity (ensuring it's a complete Markdown structure). Finally, it informs the user of the points consumed and the file location.

Scope & Considerations

• Long-Running Task: The full process (collection + generation) typically takes over 30 minutes, suitable for non-urgent, deep research scenarios. The skill proactively updates progress, but users should expect some wait time.
• Point Consumption: This is a high-cost skill; point usage scales with the number and duration of collected videos. Points for failed tasks (e.g., due to network issues) are refunded, and users can verify details in their account logs.
• Flexibility: Collection and analysis are decoupled. The same collected dataset can be reused with different analysis prompts via resume_task.py to generate reports with varying focuses, without re-collecting data.
• Dependencies & Limits: A stable network connection is required to access the server APIs. The skill does not call other external LLMs; all analysis is handled internally by Zero One Data's server. Only Douyin and WeChat Video Accounts are supported as data sources.