Generate Structured Wiki Documentation from Code

This skill acts as a documentation architect, producing structured wiki catalogues and onboarding guides directly from a codebase rather than generic templates. Its five-step procedure: scan the repository file tree and README, detect project type/languages/frameworks/architectural patterns, identify architectural layers (presentation, business logic, data access, infrastructure), generate a hierarchical JSON catalogue, and cite real files (file_path:line_number) in every section prompt rather than describing things abstractly.

The catalogue always opens, uncollapsed, with an Onboarding section built from two distinct guides: a Principal-Level Guide for senior/principal engineers - dense and opinionated, built around the single core architectural insight illustrated with pseudocode in a different language than the codebase's own, a system architecture Mermaid diagram, a domain model ER diagram, and an explicit "where to go deep" reading order with design tradeoffs; and a Zero-to-Hero Learning Path for newcomers, structured in progressive depth - Part I covers language/framework foundations with cross-language comparisons, Part II covers this specific codebase's architecture and domain model, Part III covers dev setup/testing/navigation/contributing, and appendices include a 40+ term glossary and key file reference. Beyond onboarding, the catalogue covers Getting Started (overview, setup, usage, quick reference) and a Deep Dive drilling from architecture down through subsystems, components, and methods.

Language detection picks a deliberately different comparison language to anchor the Principal Guide's pseudocode - C#/Java/Go/TypeScript codebases get compared against Python, Python codebases against JavaScript, and Rust against C++ or Go - forcing the explanation to surface what's actually distinctive about the codebase's own language choices rather than restating them in isolation. Structural constraints keep the catalogue navigable: maximum nesting depth of 4 levels, maximum 8 children per section, and small repositories (10 files or fewer) get only the Getting Started section, skipping Deep Dive, while still including the onboarding guides. Every title must derive from actual repository content - generic placeholder section names are explicitly disallowed.

Use this skill when a user asks to create a wiki, document a repo, generate documentation, understand a project's structure or architecture, get a table of contents or documentation plan, or wants an onboarding/"zero to hero" guide for a codebase.

It is specifically a codebase-documentation generator - its output is a structured catalogue with concrete file citations, not a general-purpose writing or explanation task disconnected from an actual repository to scan.

Input is a repository - its file tree, README, and source files - to document. Output is a JSON code block following the catalogue schema (items[].children[] nodes each with title, name, prompt, and children) containing the Onboarding section (Principal-Level Guide plus Zero-to-Hero path), Getting Started, and, for larger repos, a Deep Dive drilling into architecture, subsystems, components, and methods, with every prompt citing specific file_path:line_number references.

Produces Mermaid diagrams (system architecture) and ER diagrams (domain model) as part of the Principal-Level Guide, and adapts its comparison-language pseudocode based on the detected primary language of the target repository.

Teams wanting a generated wiki or onboarding guide for a codebase - serving both principal-level engineers needing a dense architectural summary and newcomers needing a progressive learning path - grounded in the actual repository content rather than generic documentation boilerplate.