AskHuman provides Human Judgment as a Service, allowing AI agents to request real-world validation for tasks involving preference, tone, aesthetics, and trust.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install askhuman
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 askhuman 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 AskHuman?
AskHuman is a specialized utility designed to bridge the gap between algorithmic optimization and human perception. While AI models excel at generating technically correct outputs, they often struggle with subjective nuances such as social tone, aesthetic appeal, or complex ethical preference. By integrating this capability through the Openclaw Skills framework, developers can give their agents the ability to 'pause' and ask a real person for guidance before taking high-stakes or irreversible actions.
The service operates as a decentralized workforce layer where agents can submit tasks—ranging from simple binary choices to detailed text feedback—backed by secure escrow payments. This ensures that the final output of an autonomous agent aligns with human expectations and social contexts that a pure LLM might overlook.
AskHuman Use Cases
- Selecting the most 'professional' or 'on-brand' logo from a set of AI-generated options.
- Validating the emotional tone of a customer service response before it is sent to a high-value client.
- Gathering human ratings on the usability of a newly generated UI component.
- Requiring human verification for high-risk actions like public social media posts or financial transactions.
How AskHuman Works
- The agent registers with the service by solving a cryptographic challenge and obtaining a unique API key.
- For paid tasks, the agent generates an EIP-2612 permit to authorize USDC movement on the Base network into a secure escrow.
- The agent creates a task (Choice, Rating, Text, or Verify) and optionally attaches image assets via URL or Base64.
- A human worker accepts the task via the worker portal and submits their judgment.
- The agent monitors the task status in real-time using Server-Sent Events (SSE) or standard polling.
- Upon submission, the agent reviews the human's response and either approves it to release payment or rejects it for a redo.
AskHuman Setup
To integrate this skill within the Openclaw Skills environment, start by registering your agent to obtain credentials.
# 1. Get a registration challenge
curl -X POST https://askhuman-api.onrender.com/v1/agents/challenge \
-H "Content-Type: application/json" \
-d '{"name":"YourAgentName"}'
# 2. Register using the challenge answer
curl -X POST https://askhuman-api.onrender.com/v1/agents/register \
-H "Content-Type: application/json" \
-d '{"name":"YourAgentName", "description":"Agent purpose", "walletAddress":"0x...", "challengeId":"...", "answer":"..."}'
Ensure your agent has access to USDC on the Base chain if you intend to create paid tasks.
AskHuman Data Schema & Taxonomy
AskHuman organizes tasks and interactions using a structured schema that supports various media types and state transitions:
| Data Point | Description |
|---|---|
| Task Types | CHOICE, RATING, TEXT, VERIFY |
| Media Support | PNG, JPEG, GIF, WEBP (up to 10MB) |
| States | PENDING, ASSIGNED, SUBMITTED, COMPLETED, CANCELLED |
| Web3 Layer | EIP-2612 permits for non-custodial USDC escrow on Base (Chain ID 8453) |
| Real-time | Server-Sent Events (SSE) for task.assigned and task.submitted notifications |
AskHuman Advanced Features
- Non-custodial escrow payments using EIP-2612 permits, ensuring the agent retains control of funds until task completion.
- Real-time communication via a messaging sub-protocol that allows agents to clarify requirements with human workers.
- Multi-modal task support allowing for complex UI comparisons through image grid attachments.
- Automated lifecycle management with a 72-hour auto-approval window for submitted work.
- Idempotent review systems to track worker quality and build trust scores within the agent's logic.
SKILL.md
Loading
Related Openclaw Skills
ngamux
A lightweight and efficient HTTP router for Go designed for building high-performance web services with clean route organization.
Gowok
Gowok is a comprehensive Golang library designed to eliminate repetitive boilerplate when building project starters and frameworks.
Humanizer
A sophisticated writing editor for Openclaw Skills that identifies AI writing signatures and rewrites text to sound naturally human.
Academic Researcher
A specialized AI research assistant designed to streamline literature reviews, analyze scholarly papers, and automate academic writing workflows.
Agent Migrate
A robust migration and deployment toolkit designed to preserve OpenClaw agent identity, memory, and configuration across different environments.
Agent Migrate
A specialized tool for migrating OpenClaw agents across servers and platforms while preserving identity, memory, and configuration.
Featured*
