cursor-council

What This Skill Does

The cursor-council skill is a powerful orchestration tool designed for developers utilizing Cursor agents within a tmux environment. It bridges the gap between individual AI coding sessions and collaborative, multi-model technical analysis. By leveraging the power of tmux sessions, this skill allows users to execute multiple coding tasks in parallel or set up a virtual 'council' of AI models, each with distinct personas and architectural philosophies, to deliberate on complex software engineering decisions.

Installation

To install the skill, use the ClawHub CLI: clawhub install openclaw/skills/skills/dongsjoa-byte/cursor-council

Ensure you have tmux installed and configured on your system, and that your cursor-agent is properly authenticated (agent login) before proceeding.

Use Cases

1. Parallel Project Execution: Accelerate development by assigning independent, non-conflicting tasks (e.g., frontend refactoring, backend error handling, and unit test generation) to separate Cursor instances running in parallel tmux sessions.
2. Architectural Deliberation: Simulate a meeting with industry luminaries by configuring different models (Opus, Sonnet, GPT) to adopt specific personas, such as architectural purists, minimalism-focused framework authors, or pragmatists. This helps in stress-testing design choices against different engineering philosophies.
3. Task Synthesis: Aggregate outputs from various agents to synthesize a comprehensive action plan for complex, multi-week feature implementation.

Example Prompts

1. 'Run the frontend and backend refactor tasks in parallel using the tmux session manager, making sure to branch off the current develop branch first.'
2. 'Set up a council debate between a system architect and a performance engineer regarding our current database schema transition, using the background documentation provided in the project folder.'
3. 'Monitor the status of all active cursor-council sessions and summarize any actionable recommendations that require human approval.'

Tips & Limitations

• Resource Management: Parallel execution is resource-intensive. On systems with 16GB of RAM, limit parallel Cursor processes to a maximum of 3-4 instances to prevent system instability.
• Dependency Handling: While agents run in parallel, ensure they work on isolated files or branches. Always verify Git status before and after mass-parallel operations to prevent merge conflicts.
• Safety Protocol: Use the --force flag with extreme caution. Always ensure your environment is clean and version-controlled. If an agent hangs, use tmux send-keys -t <session> C-c to safely interrupt the process.