praneybehl/llm-wiki-plugin — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a Claude Code plugin that turns a project folder into a living markdown knowledge base for research notes and source documents.

I want users to be able to initialize a wiki in their project, drop raw sources like papers, transcripts, notes, or PDFs into a raw folder, then ask the agent to ingest them into organized markdown pages. The wiki should keep an index, log, schema, page templates, backlinks, summaries, entity and concept pages, and should update related pages when new sources add or contradict information.

Add simple slash commands for init, ingest, query, lint, stats, graph, and upgrade. Query should answer from the existing wiki with wiki link citations, not reprocess everything from scratch. Lint should find broken links, orphan pages, missing frontmatter, oversized pages, stale claims, and suggest fixes before editing.

This should work mainly as a Claude Code plugin, but also include a portable skill that other coding agents can use with natural language. Include the helper scripts needed for search, stats, linting, and an optional graph layer. Look up current plugin and skill docs online if needed.

Want more depth? Deep Reverse