A standardized framework of five composable primitives designed to turn AI agents into persistent, self-directing autonomous workers.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-autonomy-primitives
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 agent-autonomy-primitives 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 Agent Autonomy Primitives?
Agent Autonomy Primitives provide the foundational infrastructure required to evolve AI agents from reactive chat interfaces into proactive, independent workers. By implementing this system within Openclaw Skills, developers can equip agents with the ability to manage their own backlogs, document internal reasoning, and maintain a persistent state across multiple execution cycles.
The framework is built around five core pillars: typed memory, task files, project grouping, template schemas, and heartbeat loops. Together, these primitives create a standardized language for agent operations, allowing for seamless integration between different agent architectures and human-readable interfaces like Obsidian. This ensures that every decision made and every task completed is tracked in a transparent, structured format.
Agent Autonomy Primitives Use Cases
- Establishing persistent autonomous loops for long-running AI agent operations.
- Implementing task-driven execution where agents prioritize their own work queues based on metadata.
- Building multi-agent systems that collaborate through shared memory and task vaults.
- Creating a verifiable audit trail of agent decisions and lessons learned for human review.
- Standardizing organizational workflows across different Openclaw Skills and agent frameworks.
How Agent Autonomy Primitives Works
- Initialization: The developer sets up a structured vault to act as the agent's long-term memory and workspace.
- Task Definition: Work is broken down into markdown-based task files containing YAML frontmatter for metadata like priority and status.
- Heartbeat Trigger: A periodic wake cycle (heartbeat) is established to trigger the agent's logic at regular intervals.
- Execution Cycle: During each heartbeat, the agent reads the task queue, selects the highest priority item, and performs the work.
- State Update: Upon completion or blocking, the agent updates the task status and records any new memories or lessons learned back into the vault.
- Reflection: The agent uses stored lessons and decisions to inform future execution cycles, creating a continuous loop of improvement.
Agent Autonomy Primitives Setup
To get started with Agent Autonomy Primitives, install the core CLI tool and initialize your agent vault:
npm install -g clawvault
clawvault init
You can then define your first project and task to begin the autonomous workflow:
# Create a project scope
clawvault project add "Autonomous Research" --owner my-agent
# Add an initial task
clawvault task add "Analyze market trends" --priority high --owner my-agent --project autonomous-research
Agent Autonomy Primitives Data Schema & Taxonomy
The skill organizes data into a clear directory structure using Markdown files with YAML frontmatter. This structure is optimized for both programmatic access and human readability within Openclaw Skills.
| Primitive | Directory | Metadata Fields |
|---|---|---|
| Tasks | tasks/ |
status, priority, owner, project, due, tags |
| Decisions | decisions/ |
rationale, context, date |
| Lessons | lessons/ |
experience, resolution, related_tasks |
| Projects | projects/ |
goals, owner, client, deadline |
| Templates | templates/ |
schema definitions for custom fields |
Agent Autonomy Primitives Advanced Features
- Custom Template Schemas: Modify default markdown structures to include industry-specific fields like effort, cost, or regulatory requirements.
- Multi-Agent Vault Sharing: Enable different agents to pick up tasks from the same project queue for complex, distributed workflows.
- Heartbeat-Based Autonomy: Wire primitives directly into Openclaw Skills heartbeat loops for automated periodic task processing.
- Obsidian Dashboarding: View your agent's progress, blocked tasks, and project status using native Obsidian Kanban and table views.
- Knowledge Graph Integration: Use wiki-links to create a dense network of relationships between decisions, people, and lessons learned.
SKILL.md
Loading
Related Openclaw Skills
Web Automation & Browser Workflows
A comprehensive skill for automating complex web interactions, form filling, and data extraction via a cloud-based REST API.
Web Automation & Form Interaction
A powerful automation skill that enables AI agents to fill forms, scrape data, and run browser workflows via the Apify platform.
Viral Short-Form Video Factory
A fully automated AI pipeline that researches viral trends and produces 30 platform-ready short-form videos for TikTok, Reels, and Shorts in a single run.
Video Ad Production Skill
An automated video production pipeline that converts marketing briefs into platform-optimized video advertisements using InVideo AI.
Agent Memory Templates
A production-tested framework for managing AI agent personality, memory checkpoints, and session persistence.
Bambu Lab 3D Printer Control
A local-only CLI tool enabling AI agents to monitor and control Bambu Lab 3D printers via MQTT and FTP protocols.
Featured*
