Prompt Smith

A slash command for Claude Code that optimizes rough prompts into cleaner, more executable versions.

Requires Claude Code CLI — Run this command inside a Claude Code session

---

What is Claude Code? Claude Code is Anthropic's official agentic coding CLI. Prompt Smith adds a slash command that rewrites your prompts before execution.

---

What It Does

Prompt Smith takes a rough prompt and optimizes it for Claude Code execution:

• Optimize — Rewrite for clarity, structure, and precision
• Preview — Show original and optimized side-by-side with a summary of what changed
• Choose — Pick or switch optimization modes
• Edit — Refine the optimized prompt before executing
• Confirm — Review before execution (or bypass with --yes)
• Execute — Run the optimized prompt immediately after confirmation

---

Quick Start

# 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)
claude
/prompt Refactor this function to be more readable

Update: claude plugin marketplace update prompt-smith-marketplace && claude plugin update prompt-smith@prompt-smith-marketplace

---

Command

Command	Purpose
/prompt	Optimize a rough prompt with mode selection, preview, and confirmation

Syntax

/prompt [--mode default|agentic|compact|strict] [--yes] [--dry-run] [--help] <prompt>

Flags

Flag	Purpose
--mode <mode>	Select optimization mode (default, agentic, compact, strict)
--yes	Skip confirmation and execute immediately
--dry-run	Show optimized prompt without executing (preview only)
--list-modes	Show available modes
--help	Show usage information and examples

Examples

/prompt Refactor this function to be more readable

/prompt --mode agentic Organize the work in parallel and validate each step

/prompt --mode compact Rewrite this prompt to be shorter

/prompt --mode strict --yes Fix this prompt without changing its intent

/prompt --dry-run Build a REST API with authentication

/prompt --help

Workflow

1. Parse — Extract flags and raw prompt from input
2. Select mode — Use explicit --mode or infer the best fit
3. Preview — Show mode, rationale, original, optimized prompt, and what changed
4. Confirm — Select an action: execute, edit, regenerate in another mode, or cancel
5. Execute — Run the optimized prompt in the same session

---

Optimization Modes

Mode	Purpose	Best For
default	General cleanup	Most prompts
agentic	Execution-focused structure	Orchestration, automation, multi-step delivery, repo work
compact	Shortest clean version	Utility prompts, when brevity matters
strict	Maximum fidelity	Sensitive wording, policy text, minimal rewrite

---

Sample Output

Prompt Smith
Mode: agentic
Why this mode: prompt involves multi-step execution and delivery

Available modes:
  - default   — general cleanup
  - agentic   — orchestration and execution-focused structure
  - compact   — shortest clean version
  - strict    — maximum fidelity

Original prompt:
  refactor the auth module, add tests, and make sure nothing breaks

Optimized prompt:
  ## Objective
  Refactor the auth module while preserving all existing behavior.

  ## Workflow
  1. Read and understand the current auth module structure
  2. Identify refactoring opportunities (duplication, complexity, naming)
  3. Apply changes incrementally
  4. Add or update tests to cover refactored code
  5. Run the full test suite and verify no regressions

  ## Constraints
  - Do not change the public API surface
  - Every refactored path must have test coverage

What changed: restructured into sections, added explicit workflow steps, clarified constraints

┌─────────────────────────────────────┐
│  What would you like to do?         │
│                                     │
│  > Execute                          │
│    Edit                             │
│    Regenerate                       │
│    Cancel                           │
└─────────────────────────────────────┘

---

Design Principles

Principle	Meaning
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)

---

Install

# 1. Add the marketplace (one-time)
claude plugin marketplace add maxencemeloni/claude-code-prompt-smith

# 2. Install the plugin
claude plugin install prompt-smith

Works on all platforms.

# Update (refresh marketplace first, then update)
claude plugin marketplace update prompt-smith-marketplace
claude plugin update prompt-smith@prompt-smith-marketplace

# Uninstall
claude plugin uninstall prompt-smith@prompt-smith-marketplace

---

Contributing

1. Fork the repo
2. Add/modify commands in commands/
3. Validate with claude plugin validate .
4. Submit a PR

See CHANGELOG.md for the full changelog.

---

Development

When working on Prompt Smith with Claude Code, the CLAUDE.md file at the project root provides development context including:

• Design principles — Fidelity first, Claude Code native, minimal structure
• Mode definitions — Rationale for each optimization mode
• Version management — How to bump versions and update documentation
• Local testing — How to test dev vs. plugin commands
• Test prompts — Sample prompts for verifying each mode

This context is only loaded when developing Prompt Smith itself, not when users run the command in their projects.

Release

Use /release <version> <description> to automate the full release flow.

---

More Claude Code Plugins

Plugin	Command	Description
Agent Smith	/analyze-agent	Analyzes, validates, and optimizes your Claude Code configuration. Full 7-pillar evaluation with interactive triage and guided fixes.

Check out Agent Smith for more details

---

License

MIT — see LICENSE

---

GitHub · Website