comprehension-debt

Comprehension-driven development: Claude quizzes you on your own PR diff so you catch the gap between what you think your code does and what it actually does — before you merge.

Comprehension Harness 4 Human Experiment

Agent skills that preserve human understanding and ownership during and after AI-assisted software development.

Hold an AI agent's summary against what it actually changed. Zero dependencies, one file, MIT.

A CI/CD-integrated cognitive control plane that gates pull requests behind interactive architectural quizzes to prevent developer deskilling and rubber-stamping of AI-generated code.

Measure the code nobody understands — verified-understanding loop for AI-era codebases (scan · gate · repay · prove)

Audit whether your codebase is as real as it claims. A zero-dependency static analyzer that grades every module claimed-only → reachable → tested → asserted from the code itself — never commit messages — and measures the AI-slop tax. No execution, no install of the code under audit.

Turn your Claude Code sessions into a plain-English project diary, living docs, and a visual map. One command, zero API keys.

The fathohm CLI and VS Code extension: a git-only reading of how much of a repository no human has recently written, reviewed, or explained. Generated mirror of the monorepo's MIT-licensed lanes; issues welcome here.

Adaptive programming instructor for Claude Code — structured courses, discovery-first teaching, and pluggable curricula

You merged 4,200 lines of AI code last month. How much does your team actually understand?

Always-active Claude Code plugin: Three Axes Framework — calibrates AI behavior across Mastery, Consequence, and Intent axes to prevent comprehension debt.