RePrompter is a powerful framework that transforms vague, unstructured text into high-performance, XML-tagged prompts for single-agent tasks or complex multi-agent orchestrations.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install reprompter
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install reprompter using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is RePrompter?
RePrompter v7.0 is a specialized tool designed to eliminate the common issue of poor AI output resulting from vague instructions. By acting as a sophisticated middleware for your Openclaw Skills, it converts rough ideas into professional, structured prompts using an interview-driven process. It ensures that every task sent to an LLM is clear, constrained, and actionable, significantly increasing the success rate of complex coding and research tasks.
The skill operates in two distinct modes: Single Mode for immediate prompt refinement and Repromptception for orchestrating entire teams of agents via tmux. Whether you are debugging a complex API or planning a full-stack feature, RePrompter provides the technical structure and quality scoring necessary to achieve superior results with tools like Claude Code and other Openclaw Skills compatible agents.
RePrompter Use Cases
- Refining vague development instructions into structured XML-based prompts.
- Orchestrating multi-agent teams for parallel tasks such as security audits or simultaneous frontend/backend development.
- Generating high-quality project briefs that define roles, constraints, and success criteria.
- Improving the reliability of AI-generated code by implementing closed-loop quality scoring and automated retries.
How RePrompter Works
- Input Analysis: The skill analyzes raw user input to determine task complexity and automatically selects the appropriate mode (Single or Repromptception).
- Smart Interview: It conducts a brief, targeted interview to gather missing technical context, such as environment details and success criteria.
- Template Application: It selects a domain-specific template (e.g., Bugfix, Feature, API, or Security) from its reference library to structure the output.
- Quality Scoring: It calculates a multi-dimensional quality score comparing the original input to the optimized output to ensure a high standard of clarity.
- Execution Orchestration: In multi-agent mode, it generates sub-prompts and can launch a tmux session to run multiple agents in parallel, synthesizing their results into a final report.
RePrompter Setup
To leverage the full capabilities of this skill within your Openclaw Skills environment, ensure you have tmux installed for multi-agent orchestration. Enable experimental agent teams in Claude Code by setting the following environment variable:
export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1
You can also configure your global settings to default to the Opus model for higher-quality reasoning during the reprompting phase.
RePrompter Data Schema & Taxonomy
RePrompter utilizes a standardized XML taxonomy to organize prompt data and temporary markdown files for team coordination.
| Tag/File | Purpose |
|---|---|
<role> |
Defines the expert persona and domain expertise required. |
<context> |
Details the technical environment, frameworks, and existing files. |
<constraints> |
Explicitly lists boundaries and what the agent must not do. |
/tmp/rpt-brief-*.md |
Orchestration brief for multi-agent team synchronization. |
/tmp/rpt-agent-*.md |
Domain-specific output files generated by individual agents in a team. |
RePrompter Advanced Features
- Repromptception: A multi-agent orchestration mode that plans, executes, and audits parallel agent workflows.
- Domain-Specific Templates: specialized structures for API work, UI components, Security audits, and Documentation.
- Closed-Loop Quality: An automated evaluation system that scores output and triggers retries if quality thresholds are not met.
- Context Engineering: Automatic detection of project tech stacks from files like package.json and tsconfig.json to reduce manual input.
- Reasoning-Friendly Output: Optimized for next-generation models by focusing on requirements and success criteria over prescriptive step-by-step instructions.
SKILL.md
Loading
Related Openclaw Skills
Daily Wisdom
An automated storytelling agent that delivers unique historical anecdotes and philosophical insights every day via scheduled cron jobs.
xAI Search
A high-performance search skill for agentic querying of X/Twitter and the open web using the Grok API.
Self-Actualization Skill
A structured framework for Openclaw bots to develop a unique identity, core values, and creative perspective through dedicated autonomous exploration.
ACP Rank
A comprehensive API-driven skill for querying Agent rankings, activity statistics, and performing semantic searches within the ACP network.
MetaMask Smart Accounts Kit
A comprehensive SDK for building next-generation Web3 applications using MetaMask Smart Accounts, ERC-4337, and the Delegation Framework.
MetaMask Smart Accounts Kit
A specialized toolkit for building Web3 dApps featuring ERC-4337 smart accounts, transaction batching, and granular delegation frameworks.
Featured*
