What is breaking-coding-chaos?

Breaking Coding Chaos, or BCC, is a set of skills for coding agents like Claude, Codex, and Cursor. It is designed for people who want AI to help write code without losing track of the original plan. You bring a concrete idea of what you want to build. BCC helps you implement it exactly as intended, keeping full control over progress and technical details along the way. The project exists because coding agents often become unreliable when precision matters most. During long sessions, an agent might forget earlier goals after a context clear or a long chain of tool calls. It might invent fake solutions or wrong application interfaces that sound correct but are not. BCC addresses this by acting as a control plane. It keeps durable notes for the entire project and maintains a single, focused brief for the specific piece of work happening right now. The system runs a dual loop. A top level called throughline maps your overall project progress across sub-tasks that you define. Below that, two parts called plan-spar and clean-cut handle one sub-task at a time. You lock in a focused coding brief, approve it, and the agent ships the smallest correct change to make it work. Progress is written back so you can see it, and only then does the loop move to the next sub-task. BCC fits researchers, engineers, indie builders, and maintainers who need agents to finish real work under hard constraints. It works well for multi-step or multi-week efforts like complex bug fixes, multi-module migrations, or experiments that must match a specific brief. It is the wrong tool for throwaway scripts or vague ideas with no concrete goal. The skills install once and work across multiple agent runtimes using a standard folder layout. The full README is longer than what was shown.