mcpbeat Sign in

Open Knowledge Pack Codebase Wiki Agent Skill

How to work in a Codebase Wiki project (the `codebase-wiki` starter pack): an agent-authored, source-grounded wiki of the surrounding codebase. Read when the project has a `wiki/` knowledge base with `architecture/`, `modules/`, `flows/`, `concepts/`, and `guides/` sections plus `wiki/OVERVIEW.md`, or when asked to generate or refresh a wiki of this codebase. Carries the per-folder rules and freshness + log discipline, summarizes the audience/depth knobs and source-reference convention, and bundles the full generate/refresh procedure in `references/`. Complements the platform `open-knowledge` skill; does not replace it.

5k tokens
context cost
the whole folder, loaded on every use
2
files
instructions only
0
copies elsewhere
how many repositories repackaged it
3271
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/inkeep/open-knowledge --skill open-knowledge-pack-codebase-wiki

The instruction itself

9 sections, as written by the author

Codebase Wiki pack — how to work here

This project holds an agent-authored wiki of a codebase — DeepWiki, but living in the repo. A coding agent reads the source and writes a navigable, diagram-rich, source-grounded wiki as markdown under wiki/. It is version-controlled and diffable, private by default, human+agent co-editable, renders in OK's live preview, and doubles as durable grounding context for future agent sessions. There is no separate Q&A surface — Q&A is "the OK-grounded agent + search".

> This skill is pack guidance. The platform open-knowledge skill (read/write/preview/linking/grounding rules) still governs every markdown operation — this layers the wiki workflow on top.

The shape

wiki/
  OVERVIEW.md     hub: what it is, a big-picture architecture diagram, a nav map to every section.
                  Frontmatter carries `profile` (audience/depth) + `source_commit` (freshness anchor).
  log.md          append-only generation / refresh audit trail
  architecture/   system boundaries, layers, subsystems, cross-cutting concerns + diagrams
  modules/        one page per package / module: purpose, entry points, key files, deps
  flows/          key end-to-end flows as sequence / flow diagrams + narrative
  concepts/       glossary: atomic pages for domain terms / core abstractions
  guides/         task-oriented "how / where do I change X" (filled at depth >= standard)

Generating + refreshing

Don't free-hand it — read references/generate-and-refresh.md and follow the phased, STOP-gated procedure. It auto-detects mode: a stubbed OVERVIEW.md (empty source_commit) → generate (survey → overview → architecture → modules → flows → concepts → link-graph audit); a stamped source_commitrefresh (diff source_commit..HEAD, update only affected pages, re-stamp).

Two toolsets. Read source code with NATIVE tools (Read/Grep/Glob/Bash) — OK MCP does not index non-markdown source. Author and audit the wiki with OK MCP verbs (write/edit for pages, links/search for the graph). Never hand-write wiki markdown with native Write/Edit.

The two knobs

Two natural-language knobs, read from the user's request (e.g. "build the wiki, public and exhaustive") and recorded in OVERVIEW.md frontmatter (profile: <audience>/<depth>) so refreshes stay consistent:

  • audienceinternal (default) or public. public means polished prose, no secrets / internal infra / ticket numbers, and GitHub-URL source references.
  • depthtour | standard (default) | exhaustive. Scales coverage from OVERVIEW + architecture + top flows up through per-package module pages, concepts, and task guides.

references/generate-and-refresh.md is the authoritative source for exactly how each knob shapes the output — read it before generating.

Source-reference convention

  • Intra-wiki navigation → OK doc links — they build the backlink / hub / orphan graph, so link liberally; density is how the wiki stays navigable.
  • Code references → relative links + symbol code-spans (internal) or GitHub blob URLs (public). Source-file links stay out of the navigation graph (links tracks only .md/.mdx edges, so they never show as graph dead-links or orphans) — but a wrong-depth path still surfaces in the write/edit brokenLinks response (no-such-file, or unresolvable if it overshoots the content root), so count the ../ hops from the page's folder. Never invent paths — reference only files you actually read.

The full rules — the GitHub-URL / relative fallback, the #Lxx caveat, and the exact code-span shape — live in references/generate-and-refresh.md.

Per-folder rules

architecture/ — One page per architectural area (boundaries, layers, subsystems, cross-cutting concerns). Each: a mermaid system-context or component diagram, key components (with source refs), and the design decisions behind them. Uses the architecture-page template. At depth: tour, modules fold in here.

modules/ — One page per package / module: purpose, responsibilities, public API / entry points, key files (linked per the convention), dependencies, and flows it participates in. Uses the module-page template. Skipped at tour; sub-module depth scales with the knob.

flows/ — Key end-to-end sequences as mermaid sequence / flow diagrams + narrative. Uses the flow-page template; add a Failure modes section at exhaustive. Link every module and concept the flow crosses.

concepts/ — Atomic glossary pages (one term each): definition, why it matters, where it lives in the code. Uses the concept-page template. Keep small and densely cross-linked so each concept becomes a hub.

guides/ — Task-oriented "how / where do I change X" walkthroughs: goal, steps, relevant code, gotchas. Uses the guide-page template. Populated at standard, rich at exhaustive, thin/empty at tour.

Freshness discipline (MUST)

OVERVIEW.md frontmatter carries source_commit — the git HEAD the wiki was last generated/refreshed against. It is the freshness anchor: refresh mode diffs source_commit..HEAD to update only the affected pages, then re-stamps it. Always re-stamp source_commit after a generate or refresh run — a stale anchor silently breaks incremental refresh.

Log discipline (MUST)

wiki/log.md is an append-only audit trail. Append one dated entry per generation or refresh run — one per run, not per page. Reference touched pages as markdown links (Server) so they register in the backlink graph. Entry shape:

## YYYY-MM-DD: <generate | refresh>

- Profile: <audience>/<depth>
- source_commit: <short-sha> (was <prev-sha> on refresh)
- Coverage: <sections / packages written or updated>
- Pages: [Overview](./OVERVIEW.md), [Server](./modules/server.md), ...

Templates

Each folder ships a starter template (architecture-page, module-page, flow-page, concept-page, guide-page). Create with write({ document: { path, template: "<name>" } }). Templates carry only structure (headings + frontmatter scaffold); what each section is for is described above and in references/generate-and-refresh.md, not repeated inside document bodies.

Other skills for the same job

different authors, same section of the catalogue
MCP Builder
by anthropics
vendor ×13

Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).

30k tokens scripts
Changelog Generator
by frostant
×9

Automatically creates user-facing changelogs from git commits by analyzing commit history, categorizing changes, and transforming technical commits into clear, customer-friendly release notes. Turns hours of manual changelog writing into minutes of automated generation.

774 tokens
Finishing A Development Branch
by ZhanlinCui
×7

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup

1k tokens
MCP Builder
by JayZeeDesign
×7

Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).

37k tokens scripts
Vercel React Native Skills
by vercel-labs
vendor ×6

React Native and Expo best practices for building performant mobile apps. Use when building React Native components, optimizing list performance, implementing animations, or working with native modules. Triggers on tasks involving React Native, Expo, mobile performance, or native platform APIs.

39k tokens
Vercel React Best Practices
by ratacat
×5

React and Next.js performance optimization guidelines from Vercel Engineering. This skill should be used when writing, reviewing, or refactoring React/Next.js code to ensure optimal performance patterns. Triggers on tasks involving React components, Next.js pages, data fetching, bundle optimization, or performance improvements.

34k tokens
Next Best Practices
by vercel-labs
vendor ×4

Next.js best practices - file conventions, RSC boundaries, data patterns, async APIs, metadata, error handling, route handlers, image/font optimization, bundling

20k tokens
Using Git Worktrees
by ZhanlinCui
×4

Use when starting feature work that needs isolation from current workspace or before executing implementation plans - creates isolated git worktrees with smart directory selection and safety verification

1k tokens

How to use it

Copy the folder

Take inkeep/open-knowledge-pack-codebase-wiki from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.