LLM Accumulative Obsidian Wiki

Problem

Large-model answers often produce one-off summaries. Imported articles, papers, code notes, and Q&A threads then scatter into unrelated files. For engineers, the harder problem is turning them into reusable knowledge: tracking which entities, concepts, and synthesized judgments already exist, deciding whether new evidence supports, supplements, revises, or contradicts old claims, and avoiding template pages and duplicate pages over time.

How It Works

The skill treats a Markdown / Obsidian Wiki as an accumulative artifact, not a static note folder:

• Layered structure: source/ keeps read-only raw material, wiki/ stores structured knowledge, and the root holds minimal entry points.
• Page types: sources/, entities/, concepts/, syntheses/, comparisons/, and queries/ separate single-source summaries, object facts, method concepts, cross-source judgments, confusing distinctions, and high-value answers.
• Continuous compilation: each import starts from index.md, updates existing pages when possible, and marks supports, supplements, revises, contradicts, or needs verification.
• Quality audits: batch imports, refactors, and renames run link, coverage, encoding, and thin-page checks to prevent pages that pass mechanically but remain vague.

Boundaries

It suits long-lived topic knowledge bases rather than simple one-shot Q&A. The skill favors readable English file names, concrete judgments, and cross-page references. When source evidence is weak, it still needs attributed additions, not model-generated filler.