Toolkit

Offers optional 10-15 minute learning exercises after significant work like new files, schema changes, or refactors. Uses prediction, retrieval practice, and spaced repetition so you retain what you just built rather than passively accepting the output.

Creates distinctive, production-grade frontend interfaces. Generates polished code that avoids the generic look of most AI-generated UIs. I use it when building web components and pages from scratch.

Teaches Claude Code to write proper Terraform. Covers module structure, naming conventions, state management, and common patterns so you get idiomatic IaC instead of fighting with generated code.

Scans a codebase for shallow modules, where the interface is nearly as complex as the implementation it hides, and presents the deepening opportunities as a visual HTML report. Pick one and it grills through the redesign with you, ranking each candidate by how much friction it removes.

Builds and sharpens a project's ubiquitous language. Challenges fuzzy terminology, stress-tests concepts against concrete edge cases, and checks the stated rules against the actual code. Decisions land in a glossary and ADRs as they settle rather than after the fact.

A relentless interview that pokes holes in a plan or design before you commit to it. The ADRs and glossary entries fall out of the questioning as you go, so the documentation writes itself instead of being a separate job at the end.

Spec-driven development that works with 30+ AI coding agents including Claude Code. You describe what to build, it produces a constitution, spec, plan, and task breakdown, then drives the implementation. Built-in analyze and checklist commands catch gaps before you start coding.

Token-optimized CLI proxy that sits between Claude Code and shell commands. Filters noise from tool output to cut token usage by 60-90% on dev operations like git, grep, and find. I run it system-wide via a hook so every Bash call goes through it automatically.

Python library and pair of Claude Code harnesses for AI-assisted web pentesting against your own apps. One drives HexStrike for exploratory discovery, the other runs a deterministic regression suite. Both loop fresh Claude sessions until two consecutive clean rounds.

Browser automation via the Model Context Protocol. Lets Claude navigate pages, take screenshots, click elements, fill forms, and inspect the DOM. I use it to visually review and iterate on site changes without leaving the terminal.

Connects Claude to Chrome DevTools for full browser inspection. Console logs, network requests, performance traces, Lighthouse audits, and memory snapshots. Useful for debugging and auditing sites directly from the conversation.

MCP server that gives AI agents access to 150+ cybersecurity tools for automated pentesting, vulnerability discovery, and bug bounty work. Bundles network recon, web app testing, binary analysis, and cloud security into one interface.