A meta-skill that unifies intent-engineering, dark-factory, and feedback-loop into a single self-evolving system for continuous improvement.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agentic-loop-triad
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 agentic-loop-triad 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 Unified Orchestrator v2?
Unified Orchestrator v2 is an advanced meta-skill designed to synthesize the capabilities of intent-engineering, dark-factory, and feedback-loop into a cohesive, autonomous pipeline. By bridging the gap between raw language goals and verified implementations, this tool ensures that AI agents operate within a self-correcting loop. It is a critical component for those building robust workflows within the Openclaw Skills ecosystem, providing a single entry point for complex task execution.
This system moves beyond simple task execution by introducing five major paradigm shifts, including autonomous re-specification and cryptographic improvement chains. It manages the entire lifecycle of a project—from translating goals into executable specs to measuring alignment and feeding learnings back into future iterations. By using this tool, developers can ensure their Openclaw Skills remain aligned with original intent even as complexity scales.
Unified Orchestrator v2 Use Cases
- Automating the full lifecycle from raw goal to verified code implementation.
- Detecting and correcting semantic drift in long-running AI agent tasks.
- Bootstrapping new projects by leveraging institutional knowledge from prior successful runs.
- Creating verifiable, auditable chains of improvement for AI governance and compliance.
- Managing complex workflows that require multiple specialized agents to collaborate seamlessly.
How Unified Orchestrator v2 Works
- The orchestrator translates raw natural language goals into machine-executable specifications.
- The system executes specifications autonomously while monitoring for behavioral deviations.
- Outcomes are measured against the original success criteria using a feedback loop.
- Failures trigger the generation of regression tests and automated improvement suggestions.
- The meta-learner updates the shared intent framework to make future Openclaw Skills cycles more efficient.
Unified Orchestrator v2 Setup
To get started with this meta-skill, ensure you have the core scripts accessible in your environment. You can initiate a full loop from a simple goal description using the command line:
# Run the full pipeline from a plain language goal
python scripts/pipeline.py --goal "Process customer tickets with 98% accuracy"
# To run multiple cycles for continuous improvement
python scripts/pipeline.py --goal "Optimize performance" --cycles 5 --output-dir ./run_001/
You can further customize the system behavior by editing the references/pipeline_config.json file to adjust thresholds for drift detection and meta-learning.
Unified Orchestrator v2 Data Schema & Taxonomy
The orchestrator organizes its data to ensure transparency and recoverability. Key files generated in the output directory include:
| File Name | Description |
|---|---|
pipeline_run_report.json |
The master report containing all paradigm shift outputs and logs. |
improvement_chain.json |
A cryptographically signed record of every improvement made. |
learnings_log.json |
A JSON database of cross-cycle patterns used for meta-learning. |
specification.json |
The active machine-readable intent used during the execution cycle. |
pipeline_state.json |
The state file required to resume or continue an existing loop. |
Unified Orchestrator v2 Advanced Features
- Specification Drift Detection: Automatically identifies when execution begins to wander from the original intended goal.
- Autonomous Re-specification: Adjusts impossible success criteria without human intervention when constraints are met.
- Cross-Goal Skill Transfer: Uses semantic similarity to apply learnings from one task to a completely new goal in Openclaw Skills.
- Verifiable Improvement Chains: Utilizes SHA-256 hashing to create an immutable, tamper-evident audit trail of system evolution.
- Meta-Learning Engine: Compounds learnings across different domains to improve the underlying logic of the feedback loop over time.
SKILL.md
Loading
Related Openclaw Skills
Student Tutor
An intelligent pedagogical agent that provides structured subject tutoring and heuristic problem-solving guidance for middle and high school students.
Dilraba Dilmurat Celebrity Profile
A structured data module providing comprehensive insights into the career, filmography, and biography of actress Dilraba Dilmurat.
Ed Sheeran Music Guide
A specialized music information skill providing comprehensive data on Ed Sheeran's biography, award-winning discography, and social media presence.
Work Report Generator
A powerful tool to automate the generation of structured daily and weekly work reports by analyzing git logs and project annotations.
Ideas
A lightweight and efficient way to capture, organize, and track your daily inspirations and technical ideas directly within your AI agent workspace.
AI Talking Photo
Transform any static portrait into a realistic speaking video using AI-powered facial animation and lip-syncing technology.
CardSnap API
A comprehensive API for generating, managing, and tracking professional digital business cards with integrated engagement analytics.
Featured*
