A governance layer that injects behavioral guardrails into AI agent prompts based on eight distinct operating modes to ensure security and integrity.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install moses-modes
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 moses-modes 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 Moses Governance Modes?
MO§ES™ Governance Modes is a sophisticated behavioral control system for AI agents, designed to inject specific constraints into every agent interaction. This skill acts as a critical layer in the moses-governance bundle within the Openclaw Skills ecosystem, ensuring that agents adhere to strict operational boundaries without compromising their core task instructions. By defining specific modes like high-security or high-integrity, users can prevent speculative responses or unverified transactions while maintaining agent productivity.
The system utilizes a patent-pending governance framework (Serial No. 63/877,177) that dynamically adjusts the agent's context window. This ensures that behavioral guardrails are treated as primary constraints, blocking prohibited action categories across a variety of professional and creative environments.
Moses Governance Modes Use Cases
- Restricting agents to verified data sources and preventing speculative output in high-security environments.
- Ensuring high-integrity workflows for financial or data-sensitive operations where every transaction requires verification.
- Optimizing agent performance for specific goals like research, creative brainstorming, or structured problem-solving.
- Providing a safe growth environment with self-growth constraints or enabling unrestricted exploration for developers.
How Moses Governance Modes Works
- The skill monitors the governance state by reading the active mode from the state file located at ~/.openclaw/governance/state.json.
- Before the AI agent executes any task, the skill identifies the specific behavioral constraints associated with the currently selected mode.
- These constraints are dynamically injected into the system prompt as immutable guardrails.
- The agent evaluates the task request against the active governance mode, blocking any actions that fall into prohibited categories.
- The operator can switch modes at any time using the /govern command, which triggers the moses-governance initialization script to update the system state.
Moses Governance Modes Setup
To deploy Moses Governance Modes, you must have the moses-governance skill bundle installed. You can configure the active mode via the terminal:
# Set the mode via operator command
/govern high-security
# Or directly initialize the state using the provided script
python3 ~/.openclaw/workspace/skills/moses-governance/scripts/init_state.py set --mode high-security
Moses Governance Modes Data Schema & Taxonomy
The skill manages its governance logic through a centralized state directory. The data taxonomy is organized as follows:
| Component | Path / Value | Description |
|---|---|---|
| State Directory | ~/.openclaw/governance |
The primary directory for governance metadata. |
| State File | state.json |
Stores the current active mode and behavioral parameters. |
| Primary Modes | 8 Discrete Modes | Includes high-security, high-integrity, creative, research, self-growth, problem-solving, idk, and unrestricted. |
| Dependency | moses-governance | Requires the parent governance bundle for script execution. |
Moses Governance Modes Advanced Features
- Patent-pending governance architecture for robust behavioral injection.
- Integrated state management that persists across sessions in the Openclaw Skills environment.
- Multi-mode support allowing for rapid context switching between security-hardened and creative-flexible states.
- Automated guardrail enforcement that prevents prohibited actions without manual prompt engineering.
SKILL.md
Loading
Related Openclaw Skills
Article Idea Capture
A streamlined workflow for capturing, organizing, and expanding article inspirations into structured drafts and content pools.
Feishu Sheet Tabs
A specialized skill for programmatically creating and renaming worksheet tabs within Feishu spreadsheets through intelligent browser automation.
Java Version Switcher for macOS
A specialized automation for macOS that manages multiple OpenJDK versions, handles Homebrew installations, and configures environment variables automatically.
Skill Merge and Republish
A technical workflow for consolidating redundant AI agent skills into a single, canonical version and synchronizing changes with ClawHub.
Auto Publish Created Skills
Automatically deploys and versions newly created local skills to ClawHub after successful local commits.
Ecovacs Robot Vacuum Control
An official Model Context Protocol server for managing Ecovacs DEEBOT robot vacuums through AI agents.
Featured*
