optimize-prompt-skill

optimize-prompt-skill

A Claude Code plugin + Agent Skill that rewrites prompts to reduce token consumption.

What it does

Given any LLM prompt, this plugin:

1. Identifies missing optimization dimensions (length constraints, format instructions, scope limits, negative instructions, etc.)
2. Delivers a ready-to-use rewrite immediately — no back-and-forth required
3. Estimates the expected token savings as a percentage
4. Explains what changed and why, but only if you ask

Trigger phrases

• "Optimize this prompt"
• "Reduce tokens in my prompt"
• "Make this prompt more efficient"
• "Tighten up this prompt"
• "How do I make this cheaper to run?"

Installation

Claude Code (plugin — recommended)

/plugin marketplace add https://github.com/VarunVekaria/optimize-prompt-skill
/plugin install optimize-prompt-skill@optimize-prompt-skill

Claude Code (direct install)

/plugin install https://github.com/VarunVekaria/optimize-prompt-skill

Agent Skills (Claude.ai, VS Code Copilot, etc.)

# Clone into your skills directory
git clone https://github.com/VarunVekaria/optimize-prompt-skill ~/.claude/skills/prompt-optimizer

# Or for VS Code
git clone https://github.com/VarunVekaria/optimize-prompt-skill .agents/skills/prompt-optimizer

Directory structure

optimize-prompt-skill/
├── .claude-plugin/
│   ├── plugin.json          # Claude Code plugin manifest
│   └── marketplace.json     # Marketplace catalog (enables /plugin marketplace add)
├── skills/
│   └── prompt-optimizer/
│       ├── SKILL.md         # Agent Skills entry point
│       ├── references/
│       │   └── REFERENCE.md
│       └── assets/
│           └── CHEATSHEET.md
├── README.md
└── LICENSE

Optimization dimensions

Dimension	Example addition
Length constraint	Under 150 words
Format instruction	Code only / JSON only / Bullet list
Scope limit	Assume CS background / Skip background
Negative instructions	No preamble, no caveats, no examples
Single focus	Split or drop bundled sub-questions
Persona / tone	Be terse / Be direct
Response prefill	Start the expected output structure

Example

Before:

"Can you explain how transformers work in machine learning?"

After:

"Explain how transformers work in ML. Assume CS background. Under 150 words. No preamble, no analogies, no examples unless essential."

Estimated token savings: ~55%

License

MIT — see LICENSE.