Skip to content

Ktulue/depthcheck

Repository files navigation

depthcheck

Tiered code narration for humans in an AI-first world.

depthcheck is a Claude Skill that generates layered explanations of code changes at four expertise levels — making the same work understandable to everyone from stakeholders to senior engineers.

Built live on stream at Twitch/Ktulue during Software Saturdays 04.04.26.


The Problem

AI coding tools are optimized to produce code faster. Almost none of them are optimized to help humans understand code better. The few that offer explanations do it one-size-fits-all — a single level of detail regardless of who's reading.

What depthcheck Does

Given a code change (diff, commit, PR, or session summary), depthcheck generates four tiers of explanation:

Level Audience Example
Green Stakeholders, PMs, non-technical "We added a way for users to get notified when an item they want goes on sale."
Junior Early-career devs, bootcamp grads "Created a React component that subscribes to a product's price via a REST API and sends a toast notification when it drops below the user's target."
Mid Working developers "Implemented PriceWatchService with Redis pub/sub for real-time price events, a notification queue with retry logic, and user preference storage via PostgreSQL."
Senior Architects, tech leads "Decoupled price monitoring into an event-driven microservice with fan-out via SNS/SQS, enabling horizontal scaling independent of the product catalog. Chose eventual consistency over strong consistency to keep P99 latency under 200ms."

Same work. Four audiences. Automatically generated.

Why It Matters

depthcheck isn't just a narration tool — it's a communication framework. Engineers who can explain the same work to a PM, a junior dev, and an architect are rare and valuable. depthcheck teaches that skill by example, showing what audience-appropriate vocabulary looks like at each level.

Use Cases

  • Developer Onboarding — New hires understand a codebase at their own level
  • Engineering Communication — Auto-generate stakeholder-appropriate summaries for sprint reviews, board updates, and client changelogs
  • Code Review & PRs — Tiered PR summaries help reviewers at different seniority levels understand changes faster
  • Education & Bootcamps — Level-appropriate explanations for students at different comprehension levels
  • Streaming & Content Creation — Make coding content accessible to mixed-skill audiences
  • Compliance & Audit — Document what code does for non-technical auditors in regulated industries

Getting Started

Prerequisites

  • Claude Code CLI or any Claude environment that supports skills

Installation

  1. Clone the repository:

    git clone https://github.com/ktulue/depthcheck.git
  2. Add SKILL.md to your Claude environment as a skill.

  3. Ask Claude to explain any code change:

    /depthcheck
    "Run depthcheck on this diff"
    "Give me a tiered summary of what we just built"
    "Explain this PR at all four levels"
    

Invocation

depthcheck can be invoked as a slash command (/depthcheck) or by asking Claude directly. It accepts optional arguments:

Argument Description
--hierarchical Top-down narration (Green first, then deeper)
--reverse Bottom-up narration (Senior first, then simpler)
--only <tier> Generate only one tier (e.g., --only green)
(no args) Generates all four tiers, Green through Senior

Roadmap

Stage Description Status
1. Claude Skill SKILL.md — Claude reads it and follows narration rules Complete
2. MCP Server Standalone TypeScript server for any MCP-compatible client Planned
3. GitHub Action Auto-generates tiered PR summaries on every pull request Planned
4. API / SaaS Hosted API with a dashboard for engineering teams Future

Project Structure

depthcheck/
├── SKILL.md          # The Claude Skill definition (Stage 1)
├── CLAUDE.md         # Project instructions for Claude Code
├── README.md
├── LICENSE           # MIT License
├── AI_Coding_Landscape_depthcheck_Market_Map.md
└── docs/
    └── superpowers/  # Brainstorming specs and implementation plans

Market Context

The AI coding tools market is projected to exceed $30B by 2032. Every major player (Cursor, Copilot, Devin, Sourcegraph) is focused on making AI produce code faster. Nobody is building the human-readable layer that helps different audiences understand the same work at their level. That's the gap. Full market analysis available in AI_Coding_Landscape_depthcheck_Market_Map.md.

Contributing

Contributions are welcome. Please open an issue first to discuss what you'd like to change.

License

This project is licensed under the MIT License.


Support

Buy me a coffee on Ko-fi

Created by Ktulue | The Water Father 🌊

About

A tiered narration skill that automatically explains coding sessions at four expertise levels (Green → Junior → Mid → Senior), making AI-generated code understandable to any audience.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors