llm-wiki-agent

Run Microsoft's MarkItDown as the local document extraction provider after source-extraction selects `markitdown`, or for non-wiki document-to-Markdown conversion when no intake lifecycle applies. For wiki source material, use source-extraction first so PROJECT.md provider policy, OCR/transcription approval, Source Review Gate, and the WIKI.md output contract remain in control.

Use whenever processing source material into intake Markdown, including inbox intake, document extraction, webpage extraction, OCR decisions, MinerU or MarkItDown selection, Defuddle extraction, or audio/video transcription policy checks. This skill defines source kinds, provider selection, provider safety rules, and the required output contract before Source Review Gate.

Use when reviewing this LLM Wiki development package or an initialized vault for wiki/package contract drift, process-flow closure and simplicity, init/upgrade path fit, local skill consistency, hidden or untracked residue, redundant wiki structure, conflicting agent instructions, stale workflow residue, multilingual doc drift, unnecessary branches or fallback paths, or knowledge-base cleanup risks.

Create and edit JSON Canvas files (.canvas) with nodes, edges, groups, and connections. Use when working with .canvas files, creating visual canvases, mind maps, flowcharts, or when the user mentions Canvas files in Obsidian.

Create, update, simplify, or review reusable agent skills. Use when a user wants a new skill, wants an existing skill generalized, asks whether skills are stale or overlapping, needs skill metadata/frontmatter fixed, or wants a skill packaged for reuse across Codex, LMVT, or other agent runtimes.

Create and edit Obsidian Flavored Markdown with wikilinks, embeds, callouts, properties, and other Obsidian-specific syntax. Use when working with .md files in Obsidian, or when the user mentions wikilinks, callouts, frontmatter, tags, embeds, or Obsidian notes.

Foundational Obsidian wiki static linting tool. Use as the FIRST standard layer for checking broken wikilinks and orphan pages. For project-specific business logic validations (e.g. counting files, checking specific metadata), ALWAYS run this skill first to establish a baseline, then write custom scripts for the rest. Do not write custom scripts to replace this skill's basic link parsing.

Run the Defuddle CLI as the webpage extraction provider after source-extraction selects `defuddle`, or for non-wiki webpage cleanup when no intake lifecycle applies. For wiki source material, use source-extraction first so PROJECT.md provider policy, Source Review Gate, and the WIKI.md output contract remain in control.