I-have-ADHD: A skill to stop coding agents from burying the answer
summary
AI coding assistants often bury critical information under polite filler and conversational fluff, which can be distracting and inefficient for users with ADHD. By implementing a specific set of rules that prioritize action and structure, LLMs can provide clearer, more immediate value. This approach transforms conversational responses into actionable, step-by-step instructions that reduce cognitive load. The goal is to eliminate unnecessary preamble and closers to keep the user focused on the task at hand.
key points
- The system replaces conversational fillers like 'Great question!' and 'Hope this helps!' with immediate, action-oriented answers.
- Complex tasks are broken down into numbered lists with a maximum of five items and specific time estimates to make progress manageable.
- Each response concludes with one concrete next step to prevent decision paralysis and maintain momentum.
who should read it
Developers and users with ADHD, or anyone who prefers concise, direct, and highly structured technical assistance from AI agents.
discussion skeptical
consensus
Commenters overwhelmingly agree that Claude's writing style is verbose, repetitive, and filled with artificial fillers and 'Claudisms' like 'load-bearing'. Many users feel that the model actively resists instructions to write concisely, leading to frustration and the use of external tools or strict system prompts to force readability. There is a shared sentiment that this style is a hallmark of AI slop that is becoming increasingly recognizable and annoying.
top insight
One user shares a detailed set of writing guidelines they use in an AGENTS.md file to force Claude/Fable to be legible. These rules specifically forbid em dashes, 'cataphoric teasers' like 'Here's the thing', and editorializing, while mandating ISO 24495-1:2023 plain language standards. This practical approach shifts the focus from general requests for 'conciseness' to specific linguistic constraints.
Generated by an LLM from the article and thread. Check the sources before relying on it.