wtf is fable5-mode?

fable5-mode is a Claude Code skill that tries to make a cheaper or weaker Claude model behave more like a stronger, more careful one, by forcing a strict working process rather than trying to improve the model itself. The idea behind it is that good results come from both a model's raw ability and how disciplined its working process is, and this project focuses entirely on the discipline side. Once installed, the skill asks the model to write a plan before writing any code, breaking work into small steps that each have a clear, checkable way to confirm they actually succeeded, rather than the model simply declaring a task done because it looks right. It also pushes the model to double check important output by having a separate pass try to poke holes in it, and to verify real, working results at each milestone instead of trusting that passing surface level checks means everything works. Beyond the plan itself, the project includes a set of hooks that enforce these rules mechanically instead of relying on the model choosing to follow them. These hooks can block a task from starting complex delegated work until a plan exists, prevent it from assigning work to a stronger model than allowed, remind it when it repeatedly fails the same command, and stop it from ending a task while checklist items are still unfinished. There is also a small command line checker that can be run to confirm a project's plan and progress notes are properly documented. The author is direct that this tool does not turn a weaker model into a stronger one. It closes the gap in working habits, not raw capability, and using it costs extra steps and time, so it is meant for tasks worth taking seriously rather than small quick ones. Installation is a git clone into Claude Code's skills folder, with an optional script to wire up the enforcement hooks.