agent-autonomy-primitives

What This Skill Does

The agent-autonomy-primitives skill provides the foundational building blocks for turning static AI models into persistent, self-governing agents. It introduces a structured system for memory, task management, and project organization within a local vault architecture. By utilizing specific command-line interfaces for memory types, task lifecycles, and template schemas, agents can maintain state across long-running execution loops, track dependencies, and manage their own professional backlog.

Installation

To install this skill, use the ClawHub CLI command:

clawhub install openclaw/skills/skills/g9pedro/agent-autonomy-primitives

Ensure you have initialized your vault first with clawvault init to ensure the directory structure for tasks, projects, and memory is correctly mapped to your local environment.

Use Cases

1. Autonomous Task Execution: Enable your agent to monitor a tasks/ directory, pull work, and update status fields (open to done) automatically.
2. Long-term Context Retention: Use typed memory (e.g., decision or lesson) to allow the agent to refer back to previous technical trade-offs or team preferences, preventing repetitive mistakes.
3. Multi-Agent Collaboration: Use project grouping to shard work across different agents, where each agent acts as a sub-worker for a specific project tag.
4. Self-Correction Loops: Use heartbeat loops combined with task updates to monitor for stuck processes or missed deadlines, triggering re-prioritization workflows.

Example Prompts

1. "I am assigning you a new task. Create a task file for 'Refactor Database Schema' with a high priority and set the project as 'Infrastructure-Upgrade'."
2. "Look through the project 'Backend-Refactor' and update all task statuses based on the last session's logs, then summarize any new lessons learned."
3. "Start a heartbeat loop to check the 'tasks/' folder every 30 minutes; if you see any task marked 'blocked', notify me immediately."

Tips & Limitations

• Consistent Typing: The power of this skill relies on accurate memory typing. Don't dump everything into inbox/; use decision and lesson types to ensure the agent's RAG retrieval is efficient.
• Status Flow: Always follow the defined status flow (open -> in-progress -> done) to ensure your dashboards remain accurate.
• Schema Customization: If your specific workflow requires extra fields (e.g., 'cost-estimate' or 'approver-name'), edit the YAML templates in the templates/ folder to ensure the agent strictly adheres to your custom schema.
• Limitation: The agent requires read/write access to your vault; ensure you have appropriate backups or version control (e.g., git) enabled on the vault folder to track changes made by the autonomous agent.