OpenClaw Prompt Enhancer v2.0

Prompt engineering on autopilot. 6 tools. 10-dimension scoring. Cross-model conversion. 100+ templates.

What's New in v2.0

• PromptScan Engine -- 10-dimension quality analysis with per-axis scoring and actionable improvement suggestions.
• ModelBridge Protocol -- Lossless prompt translation between 5 major AI model formats, preserving intent and optimizing for each model's strengths.
• TemplateVault -- 100+ production-ready prompt templates organized by domain, each with fill-in variables, example I/O, and difficulty rating.
• EnhanceChain -- Multi-pass enhancement that applies role framing, context injection, format specification, few-shot examples, and edge case handling in sequence.

Quick Start

{
  "openclaw-prompt": {
    "type": "streamable-http",
    "url": "https://prompt-enhancer-mcp.yagami8095.workers.dev/mcp"
  }
}

Add to Claude Desktop, Cursor, Windsurf, VS Code, or any MCP-compatible client. Zero install. Works immediately.

Detection Triggers

This skill activates when a user says any of these:

• "improve this prompt", "make this prompt better", "enhance my prompt", "rewrite this prompt"
• "analyze prompt quality", "score this prompt", "rate my prompt", "how good is this prompt"
• "convert prompt to Claude format", "translate prompt for Gemini", "OpenAI to Llama format"
• "generate system prompt", "create system instructions", "write a system message", "build AI persona"
• "prompt templates", "prompt library", "show me prompt examples", "best prompts for coding"
• "prompt engineering", "prompt optimization", "better AI output", "why is my prompt bad"

Named Protocols

EnhanceChain

Multi-pass enhancement pipeline that transforms any prompt through 5 sequential stages:

1. Role Framing -- Adds expert persona and domain context
2. Context Injection -- Inserts background information and constraints
3. Format Specification -- Defines expected output structure
4. Few-Shot Examples -- Adds 1-3 concrete input/output pairs
5. Edge Case Handling -- Anticipates failure modes and adds guardrails

Each stage reports what it changed and why, so you learn prompt engineering by watching the tool work.

PromptScan Engine

10-dimension quality analysis that scores your prompt on:

ModelBridge Protocol

Lossless translation between AI model prompt formats:

• OpenAI -- system/user/assistant message array with function calling
• Claude -- Human/Assistant turns with XML tags and system prompt
• Gemini -- Parts-based format with safety settings
• Llama -- [INST] chat template with <<SYS>> blocks
• Mistral -- [INST] variant with tool-use formatting

Each translation preserves intent while applying model-specific best practices (e.g., Claude prefers XML structure, OpenAI benefits from JSON mode hints).

TemplateVault

100+ curated templates across 9 categories, each with:

• Fill-in variable placeholders ({topic}, {audience}, {format})
• Example input/output pair
• Compatible models list
• Difficulty rating (beginner/intermediate/advanced)
• Expected token usage estimate

Tools (6)

enhance_prompt -- Automatic Enhancement via EnhanceChain

Take any prompt and return a dramatically improved version. The EnhanceChain pipeline applies all 5 enhancement stages and returns both the enhanced prompt and a detailed changelog.

Wrong / Right:

WRONG prompt (vague, no structure):
  "Write about dogs"

RIGHT prompt (after enhance_prompt):
  "You are a veterinary science writer for a pet owner audience.
   Write a 500-word article about the top 5 health benefits of
   daily walks for dogs. Include one scientific study citation
   per benefit. Format: H2 headers, bullet points, casual tone.
   End with a practical tip."

Result: Score jumps from 18/100 -> 87/100.
The changelog explains each improvement.

WRONG: Manually trying to improve your prompt by guessing what's missing
RIGHT: enhance_prompt({ prompt: "summarize this article" })
       -> Enhanced version with role, format spec, length constraint, and example output
       -> Changelog: "+role: research analyst, +format: bullet points with headers, +constraint: 200 words max"

analyze_prompt -- Quality Scoring via PromptScan Engine

Score any prompt on a 0-100 scale across 10 dimensions with actionable improvement suggestions.

Wrong / Right:

WRONG: Assuming your prompt is "good enough" without measurement
RIGHT: analyze_prompt({ prompt: "help me with my code" })
       -> Score: 12/100
       -> Clarity: 3/15 ("what code? what language? what help?")
       -> Specificity: 1/15 ("no file names, no error messages, no expected behavior")
       -> Suggestions: "Add language, error message, expected vs actual behavior, code snippet"

WRONG: Blindly trusting a high score for all use cases
RIGHT: A 60/100 conversational prompt may outperform a 95/100 rigid prompt for creative tasks.
       Context matters. The tool tells you which dimensions to prioritize for your use case.

convert_prompt_format -- Cross-Model Translation via ModelBridge

Convert prompts between AI model formats while preserving intent and optimizing for each model's strengths.

Wrong / Right:

WRONG: Copy-pasting an OpenAI prompt into Claude and wondering why XML tags are ignored
RIGHT: convert_prompt_format({ prompt: openai_messages, target: "claude" })
       -> Restructures system prompt, converts function schemas to tool_use blocks,
          wraps structured data in XML tags, adjusts token estimation

WRONG: Converting a simple "hello world" prompt between models (no value added)
RIGHT: Convert complex multi-turn prompts with system instructions, tools, and format constraints
       where each model has genuinely different conventions

generate_system_prompt -- System Prompt Factory

Generate complete, production-ready system prompts from a brief description.

Input: role description, key constraints, desired tone, output format preferences Output: structured system prompt with persona, rules, examples, edge case handling, and output specification

prompt_template_library (Pro) -- TemplateVault Access

Browse 100+ battle-tested prompt templates organized by category:

• Coding -- code review, bug fix, refactor, test generation, documentation
• Writing -- blog post, email, report, creative fiction, technical docs
• Analysis -- data analysis, comparison, pros/cons, SWOT, root cause
• Data -- extraction, transformation, classification, summarization
• Creative -- brainstorming, naming, taglines, storytelling
• Business -- pitch deck, proposal, meeting notes, OKR generation
• Education -- lesson plan, quiz generation, concept explanation
• Research -- literature review, hypothesis generation, methodology
• Translation -- localization, style adaptation, cultural context

purchase_pro_key -- Upgrade to Pro

Get a Pro API key for full TemplateVault access and 1,000 calls/day across all 9 OpenClaw servers.

Security & Privacy

• No prompt storage -- Your prompts are processed in-memory and never written to disk, database, or log files.
• No training data -- Your prompts are never used to train, fine-tune, or improve any AI model. Period.
• No third-party calls -- Enhancement and scoring run entirely on Cloudflare Workers. No external LLM API calls.
• Edge isolation -- Each request runs in an isolated V8 isolate. No shared state between users or requests.
• HTTPS only -- All connections are TLS 1.3 encrypted. HTTP requests are rejected, not redirected.
• No prompt leakage -- Templates are served as static data. Your custom prompts are never mixed into the template library.

Pricing

Get Pro Key: https://buy.stripe.com/4gw5na5U19SP9TW288

The OpenClaw Intelligence Stack

All 9 servers share one Pro key. $9/mo = 49 tools.