i-have-adhd: Open-Source Prompt Diet Skill for Concise, Action-First AI Coding Agents

An overview of i-have-adhd, an open-source MIT-licensed skill that suppresses AI coding assistant conversational fluff and compresses responses into actionable

tau · September 11, 2026

#i-have-adhd #AICoding #CodingAgent #OpenSource #DeveloperTools #Productivity

i-have-adhd: Open-Source Prompt Diet Skill for Concise, Action-First AI Coding Agents

A lightweight open-source skill designed to strip conversational filler from AI coding assistants and condense outputs into direct, actionable instructions has been released under the name 'i-have-adhd' (ayghri/i-have-adhd). Built for modern AI coding workflows across terminal and editor environments, the skill minimizes cognitive fatigue by cutting out greeting pleasantries and keeping the focus strictly on executable code and shell commands.

Visual presentation of i-have-adhd, an open-source prompt skill compressing AI coding assistant output into actionable steps

Image source: @DivyanshT91162 via X

Mainstream AI coding tools frequently open replies with verbose introductory preambles—such as "That's a great question! Let's take a look at your implementation..."—followed by multi-paragraph rationales and speculative commentary. For developers working inside compact terminal splits or fast-paced refactoring sessions, this conversational overhead forces unnecessary buffer scrolling and interrupts deep engineering concentration. i-have-adhd eliminates this conversational bloat, requiring the model to emit only numbered terminal commands and targeted code modifications with explicit file line targets.

Action-First Enforcement and the Four Structural Output Constraints

The central architecture of i-have-adhd is a strict behavioral enforcement layer that restructures model output into four deterministic elements.

  • Action First: All pleasantries, greetings, and editorial musings are banned; responses immediately present the concrete action the developer needs to take.
  • Numbered Execution Steps: Commands to run in the terminal, along with target files and line ranges to edit, are formatted into sequential, numbered steps to eliminate execution ambiguity.
  • State Restated: The assistant succinctly summarizes current task progress and modified state in one or two sentences before proceeding.
  • One Concrete Next Step: Rather than presenting an overwhelming menu of speculative options, the tool isolates exactly one concrete follow-up step requiring user decision or execution.

By restricting responses to these compact components, the skill substantially lowers the scroll time and cognitive friction required to interact with AI agents.

Preserving Focus and Reducing Cognitive Friction

Rather than modifying underlying foundation model weights or claiming improvements in code reasoning intelligence, i-have-adhd targets the efficiency of interface information delivery.

In fast-paced CLI environments, reading through large blocks of text to locate executable commands consumes valuable working memory. By stripping out redundant explanatory narration at the prompt layer, i-have-adhd allows engineers to stay anchored within their local codebase context and debugging flow. The result is a cleaner terminal buffer and a tighter feedback loop between prompt submission and test execution.

Licensing and Usage Considerations

The project is openly distributed under the permissive MIT License and hosted on GitHub at ayghri/i-have-adhd.

  • Exploratory Workflow Limitations: Because the skill aggressively compresses model verbosity, it is less suitable for open-ended architectural research or learning sessions where in-depth conceptual explanations and design tradeoffs are actively desired.
  • Functional Scope: As a behavioral output filter, it does not alter the fundamental reasoning power or code correctness of the underlying language model, serving instead as a pragmatic productivity filter for focused daily implementation work.

Sources