Tencent Docs Smart Page

Problem

When a request asks for a page, report, proposal, review, or shareable HTML, a single static file is often not enough. Templates may not match the intent, structure can be inconsistent, data may lack grounding, and the result may remain only a local file. This skill splits content-creation requests into three concrete paths: publishing existing HTML, template-based generation when the intent matches a scene, and general agent-based HTML creation when no template applies. It treats a shareable online link as the required end state.

How It Works

The skill first checks whether the request matches scene keywords in tags/scene.json. If it matches, it enters generation mode: a questionnaire confirms key inputs, a template is selected, data.js is produced, the page is injected, and the result is published. When the scene is clear, questionnaire fields can be preselected, but follow-up details are handled through the questionnaire rather than open-ended chat. If no scene matches, it switches to general agent mode: clarify requirements, generate HTML freely, then ask whether to publish it as a shareable link. Publishing follows the documented seven-step command sequence. Important constraints include putting conclusions on the first screen, maintaining the four-part structure of Overview / Controls / Charts / Logic, reading narrative.md and mock-data.js before writing data.js, and running validate_data.py --json before injection.

Boundaries

It is useful for share pages, weekly reports, proposals, retrospectives, data-analysis reports, and converting existing HTML into online links. It is not a replacement for a real business system, and unverified data should not be presented as fact. When no template covers the request, output quality depends on how clearly the user describes the requirement.