"I-have-ADHD" skill forces coding agents to lead with the answer, cut the fluff

A developer has published an open-source skill for AI coding assistants that reshapes how those tools respond, prioritizing the actual answer over the conversational padding LLMs tend to produce. Distributed as a Claude Code plugin on GitHub, it enforces ten output rules: put the next action first, number multi-step tasks, give concrete time estimates in minutes, cap lists at five items, and drop the preamble, recaps, and sign-offs like “Hope this helps!” The goal is to keep the useful content at the top of the response instead of buried beneath explanation.

Despite the name, the project isn’t tied to any diagnosis — it’s framed as a general formatting discipline for machine output. The author cites an adult ADHD self-help book as loose inspiration but explicitly adapted the principles for how a language model should reply rather than how a person should structure their day. Users install it via the Claude plugin marketplace and can fork the single SKILL.md rule file to tweak the behavior, then swap their copy in place of the upstream version.

The release lands as a small but pointed commentary on a broader complaint about current AI assistants: verbose, hedge-heavy answers that make users scroll past filler to find what they asked for. Licensed under MIT, the skill is essentially a prompt-engineering convention packaged for easy reuse, reflecting a growing DIY ecosystem of community plugins that patch over the default personalities of coding agents.