ADHD - a skill for agents
A Claude and Codex Agent SDK skill that runs many isolated, differently framed reasoning branches in parallel, then scores, prunes and deepens them, aimed at open-ended design and brainstorming prompts.
ADHD is a skill for coding agents that treats premature convergence in reasoning as an architectural problem rather than a prompting one. Where linear chain-of-thought anchors on whatever it says first, and tree-of-thought still shares one context across branches, ADHD spawns several isolated reasoning processes under deliberately different cognitive frames with no shared context during divergence, then runs a separate critic pass that scores, clusters, prunes weak branches and deepens the survivors.
It is built on the Claude and Codex Agent SDKs and is intended for design decisions, fuzzy debugging, naming, API surface design, strategy, and other prompts of the form "give me a few ways to." A published side-by-side comparison against a single-shot baseline on one evaluation problem is included in the repository's benchmark results.