CodeKeeperGuru

The problem

Vibe coding is creating a massive technical knowledge gap.

More and more people are touching codebases with AI-generated code, shipping things they don't understand, and trusting the AI blindly. Many don't even know the programming language the project is written in. That's not entirely bad: AI is opening the door to a lot of people who wouldn't have gotten in otherwise. But the long-term cost is real: codebases become impossible to maintain because AI tends to rewrite things, shift architecture, introduce patterns, and nobody on the team notices or understands what changed.

And it's not just about code. It's about concepts. Design patterns. Frameworks. Integrations like Redis, MongoDB, Kafka. How pieces fit together and why. Generic online courses don't teach this because they can't: they don't know your codebase.

The fix

Use the same AI to solve the problem it's creating.

CodeKeeperGuru is a set of Claude Code skills you install directly in your repository. You invoke them from the repo you're working on. The AI reads your entire codebase, identifies every technology, language, concept, integration, and framework in use, then builds a personalized learning curriculum and teaches you everything from scratch.

It treats you like you know nothing, and by the end you know enough to care for that codebase like a senior developer.

How it works

1. Install the skills in your target repository
2. Run /ckg-scan to inventory the codebase: language, frameworks, libraries, integrations, design patterns
3. Run /ckg-plan to generate a sequenced learning curriculum tailored to your knowledge level
4. Run /ckg-teach to start the first module, taught using real code from the repo, not fabricated examples
5. Run /ckg-next when ready to advance; a maintainer doc is written to the repo automatically
6. Repeat until the curriculum is complete

Progress is saved locally in ~/.codekeeperguru/<repo>/ so you can stop and resume any time. Docs are written to docs/codekeeperguru/ in the repo so the whole team benefits.

Install

Copy the skill directories into the target repository's .claude/skills/:

cp -r /path/to/codekeeperguru/skills/ckg-* your-repo/.claude/skills/

Or symlink them:

for skill in /path/to/codekeeperguru/skills/ckg-*; do
  ln -s "$skill" your-repo/.claude/skills/$(basename "$skill")
done

Skills

Skill	When to run
/ckg-scan	First. Scans the repo and produces an inventory.
/ckg-plan	After scan. Generates the learning curriculum.
/ckg-teach	Start or resume a lesson.
/ckg-next	When ready to advance to the next module.
/ckg-status	Any time. Shows overall progress.
/ckg-rescan	When the codebase has changed significantly.
/ckg-doc	Rarely. Only to regenerate a lost doc.

Knowledge levels

On first run, /ckg-teach asks your level:

• Zero: no programming background. The AI explains the language itself using this repo's code before covering any framework or pattern.
• Junior: knows the language but not this codebase or its stack.
• Mid: knows the stack, wants to understand how this specific codebase is structured and why decisions were made.

State files

File	Purpose
~/.codekeeperguru/<repo>/scan.json	Technology inventory
~/.codekeeperguru/<repo>/curriculum.md	Learning checklist
~/.codekeeperguru/<repo>/progress.json	Progress and vocabulary state
docs/codekeeperguru/*.md	Generated maintainer docs (committed to the repo)

Requirements

• Claude Code CLI installed and configured
• A local git repository to learn from

License

MIT. See LICENSE.