Skip to content

Claude Code Plugin

Optimize. Preview. Execute.

A slash command that rewrites your rough prompts into cleaner, more executable versions — before Claude acts on them.

MIT Licensed · v1.2.0 · Works on all platforms

Claude Code

What would you like to do?

> Execute

Edit

Regenerate

Cancel

Write rough prompts. Get clean execution.

One command. Four modes. Full control.

Smart Optimization

Rewrites your prompt for clarity, structure, and precision. Adds workflow steps, constraints, and objectives — without changing your intent.

Side-by-Side Preview

See your original and optimized prompt together. A clear summary shows exactly what changed and why, so you stay in control.

One-Click Execution

Happy with the result? Execute immediately. Or edit, regenerate in another mode, or cancel. You always decide what runs.

Four Optimization Modes

Default for general cleanup, Agentic for multi-step execution, Compact for brevity, Strict for maximum fidelity. Pick the right lens for each prompt.

CLI-Native Flags

Use --yes to skip confirmation, --dry-run to preview only, --mode to pick your lens. Designed for power users who think in terminal commands.

Intent Preservation

Prompt Smith never silently changes your intended outcome. Every optimization shows what changed, and nothing runs without your explicit consent.

From rough to refined in seconds

1

Write your prompt

/prompt Refactor this function to be more readable

Just type what you need. Don't worry about structure.

2

Pick a mode (or let us)

--mode agentic

Explicit mode or auto-inferred from your prompt.

3

Review the optimization

See original vs. optimized side-by-side with a change summary.

4

Execute, edit, or regenerate

You decide what happens next. Execute, refine, try another mode, or cancel.

Four modes for every situation

Each mode applies a different optimization lens to your prompt.

Default

Most popular

General cleanup — clarity, structure, precision. Works for most prompts.

Before
Fix the login bug
After
Identify and fix the authentication bug in the login flow, then verify with existing tests.

Agentic

Power users

Execution-focused — adds workflow steps, constraints, objectives. Best for multi-step tasks.

Before
Refactor auth and add tests
After
## Objective Refactor auth module... ## Workflow 1. Read current code...

Compact

Shortest clean version — strips fluff, keeps intent. Best when brevity matters.

Before
Can you please help me refactor this function to make it more readable and maintainable
After
Refactor this function for readability

Strict

Maximum fidelity — minimal rewrite, preserves exact wording. Best for sensitive text.

Before
Update the error message text
After
Update the error message text. Do not alter surrounding logic or other messages.

See the difference

Before

/prompt Refactor the auth module, add tests, and make sure nothing breaks

After — Agentic Mode

## Objective

Refactor the auth module while preserving all
existing behavior.

## Workflow

  1. 1. Read and understand the current auth module
  2. 2. Identify refactoring opportunities
  3. 3. Apply changes incrementally
  4. 4. Add or update tests for refactored code
  5. 5. Run full test suite — verify no regressions

## Constraints

  • - Do not change the public API surface
  • - Every refactored path must have test coverage
Same intent. Better structure. Claude Code executes it with precision.

Built on four principles

Every decision in Prompt Smith follows these rules.

Fidelity first

Never silently change the user's intended outcome

Claude Code native

Output reads like a terminal instruction, not generic chat

Minimal structure

Add structure only when it helps execution

Confirm by default

Never execute without consent (unless --yes)

Get started in three commands

Terminal

# Add the marketplace (one-time)

claude plugin marketplace add maxencemeloni/claude-code-prompt-smith

# Install the plugin

claude plugin install prompt-smith

# Use in any project

/prompt Refactor this function to be more readable

Works on macOS, Linux, and Windows

More Claude Code plugins

Tools built by the same developer to enhance your Claude Code experience.

Agent Smith

v2.5.0

Analyze, validate, and optimize your Claude Code configuration. 7-pillar evaluation with interactive triage and guided fixes.

/analyze-agent