UU AI Wiki
Paste the following prompt into your AI chat to install this skill:
Please install @user_37c39ed7/uu-ai-wiki according to https://skillhub.cn/install/skillhub.md.
About this skill
Problem
Turning scattered documents, notes, and meeting minutes into a queryable wiki often fails for more than one reason. Files may lack topic pages, source references, and a reliable update path; rescanning the full directory creates noise; and batch-converting PDFs or Office files can flood the workspace with intermediate artifacts. UU AI Wiki keeps an ordinary document folder as a Markdown-native knowledge base that can be opened in Obsidian or any Markdown editor, while allowing an agent to integrate new material traceably from the original sources.
How It Works
The workflow is not just summarizing files into isolated pages. It manages raw/, wiki/, and .uu-ai-wiki/ through a controlled ingestion pipeline:
- Intake and archiving: When creating a knowledge base, supported documents are first placed in raw/_inbox/, then organize-inbox archives them into raw/<topic>/ using stable Chinese topic categories, preventing temporary files from mixing directly into the formal source tree.
- Incremental scanning: Updates focus on the current intake queue. Legacy library migration may build one initial manifest, after which scans should remain scan_mode: incremental, avoiding repeated traversal of already ingested material.
- Structured ingestion: The agent creates or updates source pages, topic pages, entity pages, concept pages, method pages, rule pages, scenario pages, comparison pages, or synthesis pages, and refreshes index.md, overview.md, questions.md, knowledge-graph.md, and wiki/log.md.
- Quality gate: After pages and source backlinks are complete, validation runs; error entries must be fixed, warning entries reviewed, and mark-ingested used to create a quality credential. A non-empty raw/_inbox/ or further Wiki changes invalidate that credential and require revalidation.
Boundaries
By default, it does not install document conversion dependencies or batch-convert PDF, DOCX, PPTX, and XLSX files. The agent reads a binary document only when the user explicitly asks to understand its content. It does not create databases, vector stores, web frameworks, package projects, or Git repositories, making it suitable for engineering workflows that need traceable, low-side-effect, Markdown-native knowledge organization.
Use Cases
- Turn a local document folder into an Obsidian-ready Markdown wiki with topic folders and source references.
- Intake new meeting notes and requirement docs, then update topic pages, question indexes, and source citations.
- Run a Wiki health check to locate semantic contradictions, stale conclusions, orphan pages, and unsourced claims.
- With an authorized DingTalk Workspace CLI, import DingTalk Markdown docs into the local Wiki and save images.
Best For
- Engineers who manage project docs in Obsidian: need a queryable, traceable Markdown knowledge base from local folders.
- Product research operators maintaining docs: need to turn requirements and meeting notes into topic, concept, and question pages.
- Team knowledge admins connecting DingTalk: need to import DWS-authorized DingTalk documents into a local Wiki.
- Tech leads auditing local knowledge bases: need to check contradictions, stale content, orphan pages, and preserve sources.
Related Skills
Reads conversation-trace files to generate an animal- or mythology-based soul mirror card and Johari Window insights.
A Deling knowledge-base research workflow that clarifies intent, runs broad and vertical searches, supplements with web sources, validates diversity, and traces key claims.
A SiYuan knowledge-base management skill for double-link parent indexes, MOCs, numbered documents, tags, repo sync, and WeChat import workflows.
A structured workflow for academic literature reviews, covering multi-database search, screening, thematic synthesis, citation validation, and PDF output.