A sophisticated orchestration skill for running parallel Cursor AI agents and facilitating expert-level multi-model technical deliberations.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install cursor-council
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 cursor-council 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 Cursor Council?
Cursor Council is an advanced orchestration layer within the Openclaw Skills ecosystem designed to manage multiple Cursor AI instances simultaneously. By leveraging tmux to provide the necessary TTY environment, it enables developers to execute complex coding tasks in parallel or convene a digital council of AI models. This allows for diverse technical perspectives from models like Claude Opus, Sonnet, and GPT-4, effectively turning a single developer's prompt into a collaborative effort among specialized AI agents.
Cursor Council Use Cases
- Executing large-scale refactoring by splitting tasks into independent modules such as frontend, backend, and tests.
- Conducting deep architectural reviews by having different AI models adopt personas of legendary engineers to deliberate on technical debt.
- Running parallel feature development across multiple git branches simultaneously.
- Performing red-team security audits where one agent proposes code and another attempts to find vulnerabilities.
How Cursor Council Works
- Task Partitioning: The user or agent breaks a complex objective into non-conflicting subtasks or specific deliberation roles.
- Session Creation: The skill initializes independent tmux sessions for each agent, ensuring a stable environment for the Cursor CLI.
- Model Assignment: Specific AI models are assigned to each session based on their strengths (e.g., Opus for reasoning, Sonnet for speed).
- Parallel Dispatch: Commands are sent to sessions using tmux keys, allowing multiple agents to work concurrently without manual intervention.
- Output Synthesis: The user monitors progress through pane captures and synthesizes the outputs into a final actionable plan.
Cursor Council Setup
To use this component of Openclaw Skills, ensure you have the cursor-agent skill configured and are logged in via agent login. Install the required tmux dependency:
# For macOS users
brew install tmux
# For Linux users
sudo apt install tmux
Verify your environment by running which tmux && agent --version before starting a session.
Cursor Council Data Schema & Taxonomy
The skill organizes its deliberation logs and task outputs in a structured workspace for long-term persistence:
| Path | Description |
|---|---|
~/.openclaw/workspace/pr-review/ |
The base directory for all council-related archives. |
council-YYYY-MM-DD[-topic]/ |
A timestamped folder containing all artifacts for a specific session. |
README.md |
A summary document synthesized from the council's findings. |
council-[model]-prompt.txt |
The specific persona-driven prompt sent to an individual agent. |
council-[model]-output.txt |
The full technical response and raw logs from a specific AI model. |
Cursor Council Advanced Features
- Multi-Model Deliberation: Simultaneously coordinate Opus, Sonnet, and GPT models to leverage their unique reasoning styles.
- Persona Engineering: Integrated support for specialized personas like Joe Armstrong for concurrency or Martin Fowler for refactoring patterns.
- Red-Team Workflow: Orchestrate agents into attack/defense roles to harden code before deployment.
- Consensus Convergence: Iterative deliberation where agents review each others' outputs until a stable technical consensus is reached.
SKILL.md
Loading
Related Openclaw Skills
Cloudflare R2
A high-performance skill for automating file uploads and object management in Cloudflare R2 storage using the wrangler CLI.
ask-ai
A powerful bridge that allows AI agents to consult external assistants like ChatGPT and Claude when facing complex reasoning tasks.
VPS Maintenance & Optimization
A comprehensive suite for initializing, securing, and optimizing VPS servers through automated workflows.
Dify Knowledge Base Search
Search and retrieve contextual data from Dify datasets to power high-accuracy RAG workflows for AI agents.
Google AI Usage Monitor
An automated monitoring tool for tracking Gemini API RPM, TPM, and daily usage quotas within Google AI Studio.
Promotion Review AI Skill
An AI-powered evaluator that analyzes promotion materials and provides coaching to help professionals reach higher job levels.
Featured*
