Nuwa Skill Persona Distiller

Problem Addressed

When a team wants to turn an expert's, author's, or subject-matter persona's methods into a reusable AI Agent capability, the material is often scattered across local documents, WeChat articles, and existing Skills. Manual consolidation is slow and rarely produces a stable structure. This skill targets persona or theme distillation: it can work from a named person, or start from a fuzzy need, diagnose the right distillation target, and then produce a runnable Skill.

How It Works

It uses a three-source research workflow. First, it checks local files under E:\ and D:\, extracts .docx text with python-docx, and reads .md or .txt files. Second, it scrapes WeChat public-account content, typically by searching account name + keyword and using Python urllib and Playwright for headless browsing, JS redirects, lazy loading, and sub-article extraction when needed. Third, it checks existing Skills with skills_list and skill_view to reuse prior work.

Output follows a standard layout: SKILL.md with YAML frontmatter, references/ for source notes, scripts/ for reusable tools, and templates/ for template files. Versioning starts at v1.x for basic distillation, v2.x for web-material integration, and v3.x for practical case refinement. The result is an initial version that still needs Darwin-style evaluation and optimization before release.

Scope and Caveats

It fits scenarios with a clear persona, course, methodology, or case library, but it should not fabricate a person's views. WeChat scraping depends on cookies, browser availability, and target-site behavior, so results may vary under anti-scraping controls. The generated Skill is an early draft and should be verified, evaluated, and refined before being treated as a reliable capability.