wikillm

wikillm is a TypeScript CLI and plugin bundle for building LLM-maintained markdown knowledge bases. The npx wizard creates a vault with raw sources, compiled wiki articles, Obsidian config, host-specific CLAUDE.md and AGENTS.md schemas, optional outputs, and automation guidance. The plugin ships skills for ingest, query, lint, Obsidian CLI, Marp, schema regeneration, and onboarding across Claude Code and Codex.

• –Interactive vault scaffold: npx wikillm asks for personal/project mode, host targets, features, domain, and optional schedule before creating the vault structure
• –Host-aware schemas: CLAUDE.md and AGENTS.md are generated from the same TypeScript renderer, with command prefixes and install/update copy kept in host profiles
• –Compiled wiki workflow: ingest skills turn raw files into one-concept markdown articles with YAML frontmatter, [[wikilinks]], provenance, indices, and git commits
• –Query and lint skills: query starts from compiled wiki articles and graph traversal, while lint checks links, frontmatter, orphans, contradictions, stale claims, duplicates, and graph clusters
• –Obsidian and Marp integration: skills use Obsidian CLI for search/backlinks when available and can generate slide decks from wiki content through Marp
• –Automation guidance: the CLI writes .wikillm/automation.json prompts for Claude Desktop or Codex automations, but leaves actual scheduling to the selected host

• –Keeping Claude Code and Codex first-class without duplicating the whole workflow
• –Creating vault instructions without clobbering a project's own root CLAUDE.md or AGENTS.md
• –Making a markdown wiki behave like a maintained knowledge base instead of a loose notes folder
• –Being honest about automation: the CLI can record commands and prompts, but the host app still owns scheduling

The implementation separates pure vault planning from disk writes, centralizes host differences in profiles, and renders schema files in-process so setup is deterministic. The skills define the long-running behavior: ingest owns wiki compilation, query reads compiled articles first, lint keeps graph health visible, and automation metadata stays explicit rather than pretending the CLI installed a scheduler.

Ships as an npm CLI plus Claude Code/Codex plugin bundle. It gives agents a durable, plain-markdown knowledge workflow where expensive synthesis happens once during ingest, then later questions read the compiled wiki directly.