blinkingsun/claude-team-skill

This project is a skill you add to Claude Code, the command line coding assistant from Anthropic. It adds a new command called /team that turns a single coding task into a small team of AI models working together instead of one model doing everything alone. Here is how a run plays out. One model, called Fable, plans the task and breaks it into smaller pieces. A second model, Grok, checks that plan before any code gets written, pointing out risks or gaps and suggesting which pieces could run at the same time. Then the actual coding work goes to Claude models split by difficulty: harder problems go to a stronger model, routine edits go to a cheaper one. Once the work is done, Grok checks it again, this time comparing the result against the original plan and flagging anything that does not match. If problems are found, the same worker gets a chance to fix them before Fable reviews the final result, merges it, and runs the real tests. The idea behind splitting the work this way is that the checking model is never the same model that wrote the code, so mistakes are less likely to slip through unnoticed. It also keeps cost down, since the auditing and verification steps run on a separate Grok subscription rather than eating into your Claude usage. Setting it up takes three steps: installing the Grok command line tool, which needs a paid X or SuperGrok subscription, installing a bridge tool that lets Claude talk to Grok, and then installing this skill itself. Once installed, you configure a roster file that lists which model does which job, and you can edit it freely to add, remove, or swap out models. This is a tool for people who already use Claude Code and want more oversight on their AI generated changes, not a general purpose product for non-coders.