Lingyi Creator Filter

The Specific Problem It Solves

Imagine you have already identified a batch of potential WeChat Video Account creators (perhaps from manual research or past interest) and you know exactly which product you want to promote and its core selling points. The challenge is: out of these candidates, whose audience and content style best align with your product? Who has the highest conversion potential for livestream commerce? Manual comparison is time-consuming and difficult to quantify.

This skill calls a remote matching service to batch analyze your provided list of candidate creator accounts and outputs a structured matching report, helping you filter down to the most suitable collaborators from your pre-selected candidates. It does not discover new creators; it focuses solely on deep comparative analysis of your chosen targets.

How the Skill Works

The core is an asynchronous, task-driven analysis process with the following key steps:

1. Information Collection and Verification: You must provide a "list of candidate creator account names" (up to 10, 2-5 recommended) and product information (product name and core selling points are required). The skill verifies account name accuracy and guides you to supplement optional parameters like price_band or decision_type to improve matching precision.
2. API Authentication and Task Estimation: Authentication is performed using the LY_API_KEY from the skill directory's config.json. Before starting, the --estimate-only flag is used to create a task and retrieve an estimated point cost (approximately 198 points per creator). This is a mandatory confirmation checkpoint; the cost is explicitly stated, and your confirmation is required to proceed.
3. Asynchronous Task Polling: After confirmation, the scripts/creator_match.py script uses the returned task_id to poll the task status via GET /creator-match/{task_id}. The critical rule is: as long as the task is not complete (non-terminal state), polling continues with progress updates—it will not pause mid-way to wait.
4. Delivering the Structured Report: Upon successful completion, the script extracts the backend-rendered Markdown report from the server response and executes a three-level fallback save. The report is directly rendered in the chat and includes:
   • Recommendation Ranking & Six-Dimensional Match Scores: Quantitative comparison of candidates.
   • Preferred Creator Collaboration Plan: Specific actionable advice.
   • Audience Profiles, Representative Videos & Content Characteristics: Supporting analysis.
   • Actual Points Used & File Path: Clear statement of consumption and report archive location.

Applicability and Key Considerations

• Clear Positioning: This is an analysis and filtering tool, not a creator discovery tool. If you don't yet have a candidate list, please manually select creators on the Video Account platform first, then use this skill.
• Strict Input Requirements: Analysis accuracy heavily depends on the accuracy of the input creator account names (a single character difference may lead to analyzing the wrong account) and the completeness of the product information. Always copy exact nicknames from the Video Account and provide detailed product selling points.
• Significant Resource Consumption: Every new task incurs point costs, and consumption is substantial, scaling with the number of candidates and their videos. Always proceed only after the --estimate-only preview and explicit user confirmation.
• Fixed Operation Workflow: The skill operates via three core flag modes: --config (fetch available enumerations), --estimate-only (create task and estimate), and --task-id (poll for execution). Do not repeatedly create tasks while one is in progress.
• Timeout Recovery: If a task times out, you can use the original --task-id to resume polling without additional point charges, though it may require significant time.