A multi-agent governance skill that enforces a strict response hierarchy and sequencing between Primary, Secondary, and Observer agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install moses-roles
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-roles 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-roles?
moses-roles is a foundational component of the MO§ES™ governance bundle designed to bring constitutional order to multi-agent environments. By defining specific responsibilities for Primary, Secondary, and Observer roles, it eliminates redundant processing and ensures that AI agents interact within a structured, logical sequence. This skill is essential for users of Openclaw Skills who require high-integrity outputs where validation and oversight are baked into the agentic workflow.
The skill operates on the principle that sequence is law. By preventing out-of-turn responses and requiring agents to consume the context of their predecessors, moses-roles transforms a chaotic group of LLMs into a disciplined team. Whether you are building complex software or managing sensitive data, this skill provides the hierarchical framework necessary for reliable AI operations.
moses-roles Use Cases
- Enforcing a lead-and-validate workflow where a Secondary agent must verify the Primary agent's code or logic.
- Implementing an automated oversight layer where an Observer agent flags risks or constitutional violations without interfering with the task.
- Preventing 'lazy' AI behavior by forcing agents to read previous outputs and provide unique, non-repetitive contributions.
- Managing complex multi-agent simulations where the order of operations is critical to the final outcome.
How moses-roles Works
- The system assigns a specific role (Primary, Secondary, or Observer) to each agent in the environment.
- Before responding, each agent injects shared governance state from a local JSON file to understand active constraints.
- The Primary agent initiates the task, utilizing full tool access to set the analytical direction.
- The Secondary agent waits for the Primary to finish, then validates and extends the response without repeating information.
- The Observer agent monitors the entire exchange, flagging inconsistencies or gaps only after both Primary and Secondary agents have completed their cycles.
- Any attempt to respond out of sequence is automatically blocked and logged as a violation.
moses-roles Setup
To utilize moses-roles effectively, it is recommended to install the full moses-governance bundle. You can configure your workspace by updating your AGENTS.md file.
# Example of manually logging a sequence violation
python3 ~/.openclaw/workspace/skills/moses-governance/scripts/audit_stub.py log --action "sequence_violation" --detail "agent responded out of turn"
Add the role definitions to your ~/.openclaw/workspace/AGENTS.md to ensure Openclaw Skills recognize the hierarchy.
moses-roles Data Schema & Taxonomy
The skill relies on a centralized governance state and specific workspace configurations to maintain the hierarchy:
| Data Element | Path | Purpose |
|---|---|---|
| Governance State | ~/.openclaw/governance/state.json |
Stores active mode, posture, and role constraints. |
| Agent Definitions | ~/.openclaw/workspace/AGENTS.md |
Defines the specific behavior and tool access for each role. |
| Audit Logs | External Script (moses-audit) | Captures violations and state changes for compliance. |
moses-roles Advanced Features
- Broadcast Override: Use
/role broadcastto temporarily suspend sequencing, allowing all agents to respond independently for brainstorming or parallel processing. - State Injection: Automatically injects active mode and posture policies from the broader MO§ES™ ecosystem into every agent prompt.
- Automated Audit Logging: Integrated hooks that trigger the
audit_stub.pyto record every role-based action or hierarchy violation. - Local HMAC Signing: Support for an optional
MOSES_OPERATOR_SECRETto provide a secure local gate for sensitive governance changes.
SKILL.md
Loading
Related Openclaw Skills
Iterative Code Review
A sophisticated code review skill that employs multiple parallel subagents to identify bugs, security flaws, and best practice violations.
Drama Generator
An automated system that transforms written scripts into professional short-form drama videos with multi-character AI voices and cinematic transitions.
AkShare Stock Market Analysis
A comprehensive quantitative analysis tool for the A-share market that retrieves stock quotes, financial reports, and sector data using the AkShare library.
AI Competitive Competitor Analyzer
An automated intelligence framework designed to generate professional-grade competitive analysis reports with high ROI.
Prospect Research
An advanced intelligence-gathering skill that builds comprehensive pre-meeting briefs on companies and decision-makers.
Buy Me a Pie
An unofficial, API-first automation skill to programmatically manage Buy Me a Pie grocery lists and items.
Featured*
