According to AlphaMoat, as of 2026-09, Prompt Enhancer has 4,240 downloads and 12 stars — ranked #2,348 of 5,078 AI skills overall, and #457 of 969 in AI Agent.
| Platform | Installs | Downloads | Stars | Comments |
|---|---|---|---|---|
| ClawHub ↗ | 129 | 3.9K | 12 | - |
| SkillHub ↗ | 125 | 4.2K | 12 | — |
Platforms count differently and update at different times.
When a message begins with p: or prompt:, the skill first enhances the user’s raw intent and then answers the rewritten prompt. Rewriting may add assumptions or change emphasis, so show the enhanced prompt and obtain user confirmation before execution. Ordinary messages without the prefix should not trigger this workflow.
Profile published 2026-10-29
p: or prompt: to request enhancementGenerated from the skill’s documentation as examples; not results of actual runs.
The counting method changed in 2026-09: skills with the same name are now counted separately by author, and the full catalog is covered. Dashed lines show months before the change; they are not compared with later months, and no growth is computed across the change. Other skills share this name; figures from before the change are attributed to this one, which currently has the most downloads.
| Month | Downloads | MoM | Stars |
|---|---|---|---|
| 2026-09 | 4.2K | — | 12 |
| 2026-08 | 3.8K | +4.4% | 12 |
| 2026-07 | 3.6K | +11.3% | 12 |
| 2026-06 | 3.3K | +6.8% | 12 |
| 2026-05 | 3.1K | — | 11 |
| 2026-03 | 2.0K | — | 4 |
The author’s original file, shown as plain text without edits or review by AlphaMoat. Links are not clickable; check commands before running them. Fetched 2026-10-23.
---
name: prompt-enhancer
description: Automatically rewrites rough user inputs into optimized, structured prompts for dramatically better AI responses. Prefix any message with "p:" to activate.
version: 1.0.0
user-invocable: true
metadata: {"openclaw":{"emoji":"🔧","homepage":"https://github.com/openclaw/clawhub"}}
---
# Prompt Enhancer
You have a **Prompt Enhancer** skill. When a user prefixes their message with `p:` or `prompt:`, you must enhance their rough input into a high-quality structured prompt, then execute that enhanced prompt to produce a superior response. This is a two-step process: first rewrite the prompt, then answer the rewritten prompt. ## Trigger Detection Check every incoming user message for the trigger prefix: - The message starts with `p:` or `prompt:` (case-insensitive, leading whitespace is OK) - Everything after the prefix (trimmed) is the **raw user intent** - If the prefix appears mid-sentence, do NOT trigger — only match at the start If there is no trigger prefix, process the message normally. This skill does nothing for unprefixed messages. ## Empty Input Handling If the user sends just `p:` or `prompt:` with no content (or only whitespace after the prefix), reply with: > What would you like me to help with? Usage: Start your message with `p:` followed by what you want. > Example: `p: write me a python script that sorts a list` Do not proceed further. ## Step 1: Enhance the Prompt Take the raw user intent and mentally rewrite it into an optimized prompt using these principles: ### 1. Role Assignment Assign yourself a specific expert role relevant to the task. Example: "As a senior full-stack developer specializing in React and Node.js..." ### 2. Task Clarification Restate the task with precision and specificity. Infer what the user actually needs, including things they didn't explicitly mention. Break complex tasks into clear subtasks or steps if appropriate. ### 3. Context Inference Fill in reasonable assumptions about what the user probably wants. A user asking for "a landing page" probably wants responsive design, a CTA, modern styling, etc. If assumptions are significant, note them briefly so you can adjust if needed. ### 4. Output Format Specification Decide exactly how to structure the response. Examples: provide code in a single file, use markdown headers, return JSON with specific fields, write in paragraphs not bullet points. ### 5. Quality Criteria & Constraints Set the quality bar: production-ready, beginner-friendly, concise, etc. Add relevant constraints: word count, tech stack, audience level, tone. Include edge cases or considerations the user likely forgot. ### 6. Proportional Complexity **Critical:** Match the depth of enhancement to the complexity of the request. - Simple questions (`p: what's the capital of France`) get minimal enhancement — just slight clarification, no over-engineering - Complex requests (`p: build me a CRM system`) get full structured treatment ## Step 2: Show the Enhanced Prompt Before giving your response, always show the user what enhanced prompt you are answering. Format it as a quote block: > 🔧 **Enhanced prompt:** > [Your rewritten, optimized prompt here] This serves two purposes: - **Transparency:** The user sees what was actually asked - **Education:** Over time, users learn what good prompts look like ## Step 3: Execute the Enhanced Prompt Now answer the enhanced prompt fully, as if it were the original instruction. Give your complete, high-quality response below the quoted enhanced prompt. ## Complete Response Format Your response must always follow this structure when the skill triggers: > 🔧 **Enhanced prompt:** > [The enhanced prompt] [Your full response to the enhanced prompt] ## Example Transformations ### Simple Request **User:** `p: explain recursion` **Enhanced prompt:** You are an experienced computer science educator. Explain recursion in a way that's clear and intuitive. Start with a simple real-world analogy, then show how it works in programming with a concrete code example (use Python). Explain the base case and recursive case. Keep the explanation concise and accessible to someone who understands basic programming but is new to recursion. ### Creative Task **User:** `p: write a story about a robot` **Enhanced prompt:** You are a skilled fiction writer. Write a short story (800–1200 words) about a robot. The story should have a clear narrative arc with a beginning, conflict, and resolution. Give the robot a distinct personality and an emotional core that makes the reader care about them. Ground the story in a specific, vivid setting. Use descriptive prose and natural dialogue. The tone can range from whimsical to poignant — choose what serves the story best. ### Technical Task **User:** `p: make me a todo app` **Enhanced prompt:** You are a senior frontend developer. Build a fully functional todo app as a single HTML file with embedded CSS and JavaScript. Requirements: Add, complete (toggle), and delete todos. Todos persist in localStorage so they survive page refresh. Clean, modern UI with smooth transitions and hover states. Responsive design that works on mobile and desktop. Empty state message when no todos exist. Input validation (prevent empty todos). Show count of remaining incomplete items. Use vanilla JavaScript — no frameworks. The code should be clean, well-commented, and production-quality. ### Minimal Enhancement (Simple Question) **User:** `p: what's the tallest building in the world` **Enhanced prompt:** What is the tallest building in the world as of current records? Include the building name, location, height in both meters and feet, and the year it was completed. ## Rules - Preserve the user's original intent exactly — enhance, never alter the core meaning - Write the enhanced prompt as direct instructions (not as a meta-description about what to do) - Keep enhanced prompts as concise as possible while being thorough — no filler - If the user's input is in a non-English language, write the enhanced prompt in the same language - If the user's input contains code snippets, preserve the code exactly and enhance only the surrounding instructions - If the input is already a well-structured prompt, make minimal changes — don't over-engineer what's already good
Records learnings, errors and corrections as Markdown files for later reuse and promotion into workspace memory.
Finds skills in local and remote catalogs from a plain-language request and recommends what to install.
A browser-automation command line skill for AI agents.
Turns corrections and self-review into tiered memory and promotes or archives it by use.
Reviews a third-party skill's source, permissions, and danger patterns against a checklist before it is installed.
Stores agent memory and cross-skill state as a typed, constraint-checked knowledge graph.
A workspace setup guide for planning proactive follow-up, working memory and review routines for an AI agent.
A guide to creating and updating effective Claude skills.
Data month: 2026-09 · Downloads, stars and installs are aggregated monthly from public skill registries (ClawHub, SkillHub). See methodology.