alexgreensh/attention-span

Make your agents talk human. ADHD-friendly output styles for Claude Code, Codex, and others. So you can pay attention, not tokens.

What it is

A collection of "output styles" for AI coding agents (primarily Claude Code). These are small markdown instruction files that change how the agent talks to you—making its replies shorter, answer-first, and easier to scan—while leaving its actual coding ability untouched.

How it works

You drop a style file into your agent's configuration. The agent then follows those formatting rules for its chat responses. The repo ships three styles:

  • Attention-kind (flagship): Answers first, uses → markers and heavy bold so you can skim just the bold text. Built for ADHD or attention fatigue.
  • Spartan: Terse, blunt, zero warmth. Maximum signal, no transitions.
  • Rundown: Opens with a TL;DR, uses ✅/🟡/⬜ checklists, and numbers next actions so you can pick by number. Built for standups and status updates.

Why it's notable

The README includes a reproducible benchmark (12 coding tasks with hidden test suites). The headline results: coding pass rates are identical with the style on vs. off (both 97%), but the output is ~43% shorter on average, and the answer appears in the first line 75% of the time (vs. 3% without). It explicitly claims it doesn't make answers better—just shorter and more readable.

Installation & portability

  • Installable as a Claude Code plugin (one command), manually via curl into ~/.claude/output-styles/, or via a /style command.
  • Also ships as user-invoked skills (/attention-kind, /spartan, /rundown, /tldr) for the Claude chat apps.
  • The style body is plain markdown. You can strip the Claude-specific YAML frontmatter with a sed command and drop it into other agents' rules files (the README gives exact commands for Devin, Codex, and Antigravity CLI).

Who it's for

Developers using Claude Code or similar agents who are tired of walls of text, want to skim answers quickly, or want a cleaner status-report format—without sacrificing the agent's coding performance.

Related

  • Dispatch
    Maximizing Claude Code Session Value and Token Efficiency

    Optimizing Claude Code sessions requires strategic use of prompt caching, context management, and subagents to reduce token costs and prevent cache misses.

  • Project
    YeonwooSung/my-claude-code-mods

    A set of Node‑based plugins (mods) for Anthropic’s Claude Code that add profiling, context analysis, timeline snapshots, loop detection, pinned notes, and model‑effort visualisation. Each mod is installable via Claude’s marketplace and includes unit tests and TypeScript type‑checking.

  • Project
    Piebald-AI/claude-code-system-prompts

    A repository that extracts, version‑tracks, and publishes every system‑prompt string used by Anthropic’s Claude Code code‑assistant, providing raw prompt files, token counts, and a changelog for over 500 fragments.

  • Dispatch
    Mastering Claude Code: From Prompting to Programmable Engineering

    A deep dive into transforming Claude Code from a simple chatbot into a sophisticated autonomous agent using custom skills, subagents, and the Model Context Protocol.