GitHub Copilot Prompt Pack + Custom Agent

Steno Mode for GitHub Copilot

Stenographer Mode — Save ~40% tokens in AI responses

Steno Mode (also called Stenographer Mode) is a GitHub Copilot prompt pack and custom agent that cuts ~40% of response overhead through structured shorthand. Works with VS Code Copilot, Claude Code, Kiro, ChatGPT, Cursor, Windsurf, Aider, and Cody. Exact code, commands, and paths always preserved.

~40%

Average token reduction from baseline to Steno across the benchmark corpus.

50+

Benchmark samples spanning code review, debugging, architecture, docs, and failure cases.

8+

AI coding assistants supported: VS Code Copilot, Claude Code, Kiro, ChatGPT, Cursor, Windsurf, Aider, Cody.

Where to save tokens

Where to save tokens in AI

VS Code Copilot + Kiro

Install /steno prompt and Steno agent with one command. Works across Ask, Edit, and Agent mode.

Claude Code + Claude

Paste packs/claude/system.txt into your Claude system prompt. Token saving activates immediately.

ChatGPT custom instructions

Drop packs/chatgpt/custom-instructions.txt into ChatGPT's custom instructions field.

Cursor · Windsurf · Aider · Cody

Paste packs/cursor/rules.txt into rules, system prompt, or agent instructions on any of these surfaces.

Install

How to use Steno Mode

Install directly from GitHub with npx --yes github:AkashAi7/stenographer-mode install --scope user.

Use --scope project for repo-local installation into .github/prompts and .github/agents.

FAQ — Token Saving

How to save tokens in AI

How to save tokens in AI?

Use a structured shorthand prompt like Steno Mode. It cuts filler and soft phrasing from AI responses while keeping exact code, commands, and paths. Install once, save ~40% tokens on every technical reply.

Where to save tokens?

Save tokens in your system prompt, custom instructions, or agent config. Steno Mode installs into VS Code Copilot with one npx command, or paste a starter pack into Claude Code, Kiro, ChatGPT, Cursor, or Windsurf.

What is meant by a token?

A token is ~3–4 characters or ~0.75 words. It is the billing unit for LLM APIs. Every token in a response costs money and time. Steno Mode reduces tokens by ~40% without losing technical content.

Does Steno Mode work with Claude Code and Kiro?

Yes. Paste packs/claude/system.txt into Claude Code, or packs/cursor/rules.txt into Kiro's agent instructions. Token saving activates immediately.