opennexum

What This Skill Does

OpenNexum is a robust, contract-driven multi-agent orchestration layer designed specifically for the OpenClaw ecosystem. It transforms chaotic agent-driven development into a structured, verifiable pipeline. At its core, the skill utilizes a "Generator-Evaluator" pattern where one agent writes code or documentation while an independent evaluator agent verifies the output against a strict YAML-based contract. By enforcing immutable contracts, it prevents drift in large projects, ensures code quality through iterative feedback loops, and automates the tracking of complex task lifecycles from pending to completion.

Installation

To integrate OpenNexum into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/ayao99315/opennexum Once installed, you must initialize your project directory using the provided script: nexum-init.sh --project my-new-app --tech "react-typescript" This creates the necessary directory structure, including the docs/nexum/contracts/ folder and the runtime tracking files required for agent coordination.

Use Cases

1. Parallel Feature Development: Coordinate multiple coding agents to work on independent modules of a large architecture without merge conflicts, as each module is constrained by a specific task contract.
2. Automated QA Loops: Deploy a generator agent to implement a feature, followed by an automated evaluator agent that runs unit tests and static analysis, automatically flagging failures and reverting changes if the criteria are not met.
3. Long-Running Project Management: Manage complex, multi-day development cycles where task dependencies must be respected. OpenNexum logs every interaction in the events.jsonl file, providing a full audit trail.

Example Prompts

1. "OpenNexum: Initialize a new contract for the authentication module with high-security requirements and assign the generator to user-auth-agent."
2. "Dispatch the evaluator for task ID: 004-auth-refactor to verify the current iteration output against the requirements defined in the contract YAML."
3. "Run a health check on the current project. Are there any stalled generator agents that need a manual status update?"

Tips & Limitations

• Tip: Always define your eval_strategy clearly in the contract. The more specific your YAML schema, the more effective your evaluator will be in identifying edge cases.
• Tip: Use nexum-revert.sh for any task marked as 'failed' to cleanly roll back the working directory before restarting the iteration.
• Limitation: Contracts are immutable. If a requirement changes significantly during development, you must create a new task ID rather than attempting to patch an existing contract.
• Limitation: The system requires a stable environment; ensure your active-tasks.json has read/write permissions to avoid locking issues during parallel agent execution.