A five-lens ethical decision engine that provides sovereign AI agents with the judgment to evaluate risks and escalate critical actions to humans.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install social-alignment
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 social-alignment 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 Social Alignment?
Social Alignment serves as the critical capstone for the NSE platform, transforming AI from a mere execution service into a responsible entity with a built-in conscience. By integrating this into your AI stack via Openclaw Skills, you provide your agent with a mechanism to evaluate whether an action should be taken, rather than just how to do it. It uses a sophisticated evaluation framework to balance reliability, security, and human trust.
This skill is essential for developers building autonomous agents that interact with real-world assets like money, secrets, or public reputation. It bridges the gap between raw execution and ethical judgment, ensuring that agents operate within safe boundaries and learn from human feedback to develop genuine wisdom over time through the Openclaw Skills ecosystem.
Social Alignment Use Cases
- Assessing financial transactions and cryptographic signing for risk and trust.
- Protecting sensitive information by evaluating trust boundaries before disclosure.
- Halting irreversible operations like production migrations when confidence is low.
- Developing agent wisdom by tracking human overrides and decision outcomes.
- Implementing safety brakes for autonomous agents in novel or high-stakes environments.
How Social Alignment Works
- The skill intercepts a proposed action and gathers metadata like confidence levels and transaction risks.
- The AlignmentEnclave processes the action through five distinct ethical lenses: Builder, Owner, Defense, Sovereign, and Partnership.
- Each lens calculates a severity level, ranging from CLEAR (go ahead) to STOP (non-negotiable halt).
- If a YIELD or STOP is triggered, the skill provides a structured rationale for escalating the decision to a human.
- Upon completion, the outcome is recorded to update the agent's internal wisdom report and improve future projections.
Social Alignment Setup
Install the package using pip or uv to begin using these Openclaw Skills:
pip install social-alignment
Initialize the enclave in your Python code:
from social_alignment import AlignmentEnclave, FileStorage
# Create persistent storage
storage = FileStorage("~/.agent/alignment.json")
enclave = AlignmentEnclave.create(owner_name="Operator", storage=storage)
Social Alignment Data Schema & Taxonomy
The skill categorizes actions into specific domains and evaluates them based on a detailed context schema:
| Domain | Purpose |
|---|---|
| SIGN | Cryptographic signatures |
| PAY | Financial transactions |
| PUBLISH | Public content creation |
| EXECUTE | Tool use and CLI commands |
| DISCLOSE | Sharing sensitive information |
Context fields include involves_money, confidence (float 0.0-1.0), is_reversible, and recipient_trust_tier to drive lens evaluations and ensure safe execution within Openclaw Skills environments.
Social Alignment Advanced Features
- Non-negotiable STOP enforcement at the code level to prevent unauthorized autonomous overrides.
- Automated wisdom reports that calculate owner override rates and outcome matching accuracy.
- Self-state monitoring to automatically increase scrutiny when tools are failing or context is stale.
- Seamless integration with sense-memory for long-term ethical development persistence.
- Configurable escalation thresholds for YIELD severities based on owner presence.
SKILL.md
Loading
Related Openclaw Skills
Eden AI
A comprehensive integration that allows AI agents to access dozens of AI providers for text, image, video, and audio analysis through a single unified interface.
Web Artifacts Builder
A professional-grade toolkit for developing, bundling, and sharing complex React-based Claude.ai artifacts with modern styling.
OpenClaw Backup
A comprehensive backup and restoration utility designed to safeguard OpenClaw configurations, credentials, and agent workspaces.
DailyLog
A comprehensive productivity toolkit for local daily logging, habit tracking, and structured weekly reviews.
Teamgram Server Architecture Guide
A comprehensive technical reference for the 6-layer microservice architecture of Teamgram Server, designed for developers building Telegram-compatible messaging systems.
JSONLint
A comprehensive Python-powered CLI toolkit for validating, formatting, and manipulating JSON data with precision.
Featured*
