Lattice is a file-based operating system that enables AI agent teams to execute complex, long-running projects through a structured 8-phase pipeline.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install lattice
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 lattice 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 Lattice?
Lattice transforms volatile AI chat interactions into a persistent, file-driven operating system for agent teams. By replacing ephemeral sessions with a structured file state, it ensures that Openclaw Skills can handle development cycles lasting hours or days without losing context or drifting from the original goal. This framework provides the stability needed for professional-grade AI engineering.
At its core, Lattice utilizes a sophisticated 8-phase pipeline that separates high-reasoning tasks like architecture and review from token-intensive tasks like implementation and testing. This allows developers to optimize costs by assigning the most appropriate LLM to each specific stage of the workflow while maintaining a rigorous three-tier failure handling system to unblock agents automatically without human intervention.
Lattice Use Cases
- Orchestrating complex software engineering projects that exceed standard context windows.
- Managing multi-department AI organizations with specialized agent roles.
- Reducing token costs by mapping specific model strengths to pipeline phases.
- Automating long-running research and development tasks via cron-scheduled Openclaw Skills.
- Implementing autonomous failure recovery through model escalation and peer consultation.
How Lattice Works
- The user initializes an organization structure defining departments and projects.
- A project-specific 8-phase pipeline (from Constitute to Gap Analysis) is established.
- A cron-scheduled orchestrator monitors the pipeline state files to trigger the next logical step.
- If an agent encounters an error, the system automatically escalates through stronger models or consults peer agents for a resolution.
- All progress is persisted in Markdown and JSON files, allowing the agent team to resume work across sessions seamlessly without context loss.
Lattice Setup
To begin using Lattice within your collection of Openclaw Skills, run the initialization command:
# Initialize a new AI organization
lattice init
Follow the interactive prompts to define your organization name, target directory, and initial departments. You will also need to configure your PIPELINE_STATE.json with your preferred model escalation chain and set up the orchestrator cron job to drive the automation autonomously.
Lattice Data Schema & Taxonomy
Lattice uses a hierarchical file structure to maintain state and history across Openclaw Skills workflows:
| File or Folder | Description |
|---|---|
| ORG_README.md | Central hub for organizational structure and departments. |
| TASKBOARD.md | Global priority list and task tracking for the entire organization. |
| PROJECTS/PROJECT_NAME/STATUS.md | Human-readable progress and current phase status for a project. |
| PROJECTS/PROJECT_NAME/PIPELINE_STATE.json | The core machine-readable state machine for the 8-phase pipeline. |
| PROJECTS/PROJECT_NAME/DECISIONS.md | Log of architectural choices and rationale made by agents. |
| DEPARTMENTS/DEPT_NAME/CHARTER.md | Mission statement and operational boundaries for specific teams. |
Lattice Advanced Features
- Three-tier failure handling including Model Escalation, Peer Consult, and Auto-Triage.
- Per-phase model configuration to optimize token spend vs. reasoning power.
- Multi-project parallel execution with independent cron-scheduled orchestrators.
- Automated handoff protocols between departments for complex cross-functional workflows.
- File-driven state machine that prevents context window overflow and agent drift.
SKILL.md
Loading
Related Openclaw Skills
Ollama Web Search
A powerful web search integration for AI agents utilizing the Ollama API to fetch real-time information.
Daolv Hotel Search
A specialized AI agent skill for rapid hotel discovery, filtering, and shortlisting based on budget, location, and amenities.
SEO Affiliate Guide Expert
A specialized AI framework for generating SEO-optimized, high-converting affiliate purchase guides and product reviews in Spanish.
PDF Tools
A comprehensive suite of Python-based utilities for viewing, extracting, and editing PDF files.
Sats4AI
An MCP server providing a decentralized marketplace for AI tools where users pay for exactly what they use via Bitcoin Lightning micropayments.
unbrowse
A drop-in browser replacement for AI agents that turns web traffic into high-performance, reusable API endpoints.
METADATA
Featured*
