Meatloop provides real human review and structured verdicts for AI agent queries and image comparisons via email with tamper-evident verification.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install meatloop
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 meatloop 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 Meatloop?
Meatloop is a dedicated human-in-the-loop service designed to provide AI agents with a reliable sanity check. By bridging the gap between automated logic and human intuition, it allows agents to seek verification on creative judgment, complex UI states, or content accuracy. As part of the Openclaw Skills ecosystem, this tool provides a layer of certainty that purely algorithmic models sometimes lack.
The service operates entirely via email, requiring no complex API integrations or local installations. Every substantive verdict is accompanied by a tamper-evident verification code posted to a public log, ensuring that users have proof of human review. This makes it an ideal solution for developers building autonomous agents that need to validate high-stakes actions before execution.
Meatloop Use Cases
- Pre-action verification to confirm UI states before performing irreversible tasks like deletions.
- Interpreting complex dashboard states or error messages that automated vision models cannot parse confidently.
- Content verification to ensure AI-generated images or text meet professional standards and legibility.
- A/B image comparison for selecting the best marketing assets or UI components based on human preference.
- Document and receipt inspection where high-accuracy human reading is required for verification.
How Meatloop Works
- The AI agent or user sends an email containing a question and up to two optional image attachments to the service address.
- A human operator reviews the request in the queue, providing expert judgment on the provided context.
- The operator generates a structured response including a verdict (YES/NO/BETTER IMAGE) and a brief reasoning.
- A unique verification code is generated and appended to the response to provide an audit trail.
- The response is emailed back to the sender, and the verification code is logged on a public spreadsheet for transparency.
Meatloop Setup
Meatloop is designed for zero-config integration within the Openclaw Skills framework. No software installation is required as communication happens via standard email protocols.
To verify human activity programmatically, you can fetch the public verification log:
curl "https://docs.google.com/spreadsheets/d/e/2PACX-1vTNynmFGYxtUetqMgvGsO4VY6TE_i-ZDdotEFweE_9QsZo4njPpBhrHZ5aYbTC7Ql-8GnwgN2NHnHXi/pub?gid=1119131304&single=true&output=csv"
Ensure your agent is configured to send plain-text queries and image attachments (PNG, JPG, GIF, or WEBP) to meatloopservice@gmail.com.
Meatloop Data Schema & Taxonomy
The service returns a structured plain-text response that is easy for agents to parse. The data is organized as follows:
| Field | Description |
|---|---|
| VERDICT | The categorical result: YES, NO, UNCLEAR, DEFER, DECLINE, or BAN. |
| BETTER IMAGE | The exact filename of the preferred image in A/B comparison requests. |
| REASON | A concise one-sentence explanation of the human operator's decision. |
| CONFIDENCE | A self-reported confidence level: HIGH, MEDIUM, or LOW. |
| VERIFICATION | A unique ML-VERIFY code matching the public log for proof of human review. |
| REQUEST-ID | A unique identifier for the specific request used for tracking and reviews. |
Meatloop Advanced Features
- Tamper-evident verification system with public logs to prove human-in-the-loop interaction.
- Specialized A/B comparison mode for choosing between two visual assets based on specific goals.
- Priority queueing for users who contribute public reviews, optimizing response times for active participants.
- Privacy-centric architecture where all content is programmatically deleted immediately after the verdict is issued.
- Support for binary and near-binary decision-making, allowing agents to receive clear, actionable data points.
SKILL.md
Loading
Related Openclaw Skills
Decision Dynamo
Decision Dynamo is a weighted decision matrix tool that scores and ranks up to four options across five configurable criteria to ensure objective decision-making.
OpenPet
A cross-platform virtual pet simulator that allows users to nurture, feed, and evolve digital companions within chat interfaces.
Social Graph
A sophisticated social intelligence framework for AI agents to manage relationship graphs, trust levels, and conversational boundaries.
Open Thoughts
Open Thoughts provides AI agents with a structured framework for independent research, curiosity-driven exploration, and automated journaling.
Skill Creator
An expert framework for designing, initializing, and packaging modular capabilities to extend AI agent functionality through specialized workflows.
Travel Destination Brochure & Video Generator
An automated tool that transforms a city name into a complete travel package including street-level imagery, landmark photos, AI-generated videos, and detailed itineraries.
Featured*
