A skill that helps Claude Code users compress prompts and cut token usage while keeping output quality high. It covers compression patterns, system prompt templates, XML tag structuring and one-shot examples.
Claude Code
$ npx -y skills add samibajwaisking/Token-Optimizer-Skill --agent claude-code
Codex
$ npx skills add samibajwaisking/Token-Optimizer-Skill -a codex
Cursor
$ npx skills add samibajwaisking/Token-Optimizer-Skill -a cursor
Gemini CLI
$ npx skills add samibajwaisking/Token-Optimizer-Skill -a gemini-cli
GitHub Copilot
$ npx skills add samibajwaisking/Token-Optimizer-Skill -a github-copilot
OpenCode
$ npx skills add samibajwaisking/Token-Optimizer-Skill -a opencode
Windsurf
$ npx skills add samibajwaisking/Token-Optimizer-Skill -a windsurf
⭐ Star this repo to unlock usage rights
Built by Sami Bajwa — AI Educator & Automation Builder
Token Optimizer is a production-ready Claude AI Skill that teaches you how to extract the highest-quality, most complete responses from Claude using the fewest possible input tokens.
💡 Most people waste 60–80% of their tokens on filler, bad formatting, and unstructured prompts — and get worse results because of it. This skill fixes that.
A fully optimized session uses 3–8x fewer tokens than an unoptimized one — for the same or better output quality.
Turn 50-token verbose prompts into 8-token precision instructions. Shorthand syntax for every common task type.
Eliminate 30–50% of response bloat. Tell Claude exactly what format, length, and structure you want.
The most underused prompting technique. Reduce clarification rounds to zero with structured input.
Bundle 4 messages into 1 structured prompt. Stop the back-and-forth.
Tell Claude what NOT to do. Cuts filler by 20–35% instantly.
One perfect example replaces 100 tokens of style instructions. Ready-made examples for 10 domains included.
token-optimizer-skill/
│
├── SKILL.md # 🧠 Core skill file — load this into Claude
│
├── references/
│ ├── compression-patterns.md # 50+ shorthand patterns across 6 categories
│ ├── output-formats.md # 10 format templates + Anti-Filler Block
│ ├── system-prompt-templates.md # 10 production-ready system prompts (<200 tokens each)
│ ├── one-shot-examples.md # Ready-made examples for 10 domains
│ └── context-management.md # 7 techniques for conversation pruning
│
├── examples/
│ └── before-after.md # 20 real prompts with token counts (avg 65% savings)
│
├── .github/workflows/
│ └── star-reminder.yml # Auto-reminds contributors to star
│
├── LICENSE # Star-Required License
├── CONTRIBUTING.md # How to contribute
├── CHANGELOG.md # Version history
└── README.md # This file
| Technique | Average Token Savings |
|---|---|
| Prompt Compression | 40–70% input reduction |
| Output Format Control | 30–50% response reduction |
| XML Structuring | Eliminates 2–3 clarification rounds |
| Instruction Bundling | 60–80% fewer messages |
| Negative Instructions | 20–35% shorter responses |
| One-Shot Examples | Replaces ~100 tokens of style description |
| Context Pruning | 70–90% history token reduction |
| Combined Effect | 3–8x fewer tokens. Same quality. |
SKILL.md as a project instruction fileSKILL.mdCopy the full contents of SKILL.md → paste into Claude's system prompt field
with open("SKILL.md", "r") as f:
system_prompt = f.read()
# Pass as system_prompt in your Anthropic API call
response = client.messages.create(
model="claude-sonnet-4-20250514",
max_tokens=1024,
system=system_prompt,
messages=[{"role": "user", "content": "Your task here"}]
)
"Optimize this prompt for me: [your verbose prompt]"
"Help me write a system prompt for a customer support bot"
"Show me the compressed version of this instruction"
"How do I structure this multi-part request efficiently?"
"Reduce tokens in this conversation history"
"Write a one-shot example for [my use case]"
The fastest way to improve any prompt immediately:
[ROLE (1 line)] + [TASK (compressed)] + [FORMAT (explicit)] + [CONSTRAINTS (negative)]
Example — Before (74 tokens):
I have a Python script that I wrote and it's having some issues. Can you please
review it carefully and tell me what's wrong with it and also suggest how I could
improve it to make it better and more efficient?
After (12 tokens):
You are a Python expert. debug+optimize: [code]
Format: issues list → fixed code. No intro.
Same output. 84% fewer tokens.
| File | Contents |
|---|---|
compression-patterns.md | 50+ shorthand patterns: analysis, writing, code, extraction, research, ideation |
output-formats.md | 10 format templates + Universal Anti-Filler Block + length modifiers |
system-prompt-templates.md | Code assistant, content writer, data analyst, support bot, researcher, summarizer, translator, creative, tutor, business automation |
one-shot-examples.md | Code review, email, data extraction, simplification, summaries, social media, meeting notes, product descriptions, SEO, feedback |
context-management.md | State Summary Method, Rolling Compression, Stateless Design, Role Pinning, Reference Injection, Checkpoint Commits, Context Scoping |
This repository uses a Star-Required License.
⭐ You must star this repository before using, copying, or distributing this skill. Attribution to Sami Bajwa / samioutic.com must appear in any product that uses this skill.
See LICENSE for full terms.
⭐ Click here to star and unlock usage →
Sami Bajwa is an AI Educator, web developer, and digital entrepreneur based in Sargodha, Pakistan. He operates under the brand SAMIOUTIC (samioutic.com) and is dedicated to helping Pakistani freelancers and learners master AI tools and earn online.
| Project | Description |
|---|---|
| Samioutic.com | Personal brand — AI education, web development, digital services |
| CIPHER Trading Skill | Claude skill for institutional-grade crypto/forex trading signals |
| HabitQuest Pro | Gamified habit tracking system (GitHub Pages) |
| VoiceScribe AI | Bilingual Urdu/English voice-to-text prototype |
| Universal AI Chat | Single-file app supporting 8 AI providers (Claude, GPT, Gemini, Groq...) |
Contributions that improve token efficiency are welcome. Before contributing:
| Version | Date | Description |
|---|---|---|
| v1.0.0 (Latest) | 2025-05-28 | Initial release — 5 pillars, 6 reference files, 20 before/after examples |
See CHANGELOG.md for full version history and roadmap.
If this skill saves you tokens and time — drop a ⭐ on the repo.
It helps other developers and AI users find it.
Built by Sami Bajwa · Samioutic · Pakistan 🇵🇰
Stars on GitHub
A skill that shapes Claude Code's output for readers with ADHD: it leads with the next action, numbers multi-step work, restates state across turns, drops tangents, gives specific time estimates, and makes progress visible. Invoke it with /i-h.
@ayghriGives your AI agent working knowledge of Obsidian, so it can read and write notes, build Bases database views and JSON Canvas mind maps, and turn web pages into clean Markdown for your vault.
@kepanoGives product managers step-by-step help with discovery, prioritization, strategy, launch and growth work, so they get structured outputs instead of loose text. Also includes skills for reviewing AI-built code before shipping.
@phurynFormerly: Colleague Skill / colleague-skill.
@titanwingsA skill that helps Claude Code users compress prompts and cut token usage while keeping output quality high. It covers compression patterns, system prompt templates, XML tag structuring and one-shot examples. Install it with the command on its page. It includes Token-Optimizer-Skill. Its skills do not fire on their own yet. Free and open source.
Is this plugin yours?
Claim it with GitHubSubmit a pluginPromote it
