autonomous-skill

What This Skill Does

The Autonomous Skill is an advanced task management and execution framework designed for complex, multi-stage projects within OpenClaw. It leverages a dual-agent architecture (Initializer and Executor) to bridge the gap between initial intent and long-term execution. When a user defines a project, the skill creates a dedicated directory structure in .autonomous/, where it maintains a task_list.md and progress.md. This allows the agent to maintain state, context, and history across multiple sessions, ensuring that progress is never lost, even if a task takes hours or days to complete. The skill uses Claude Code headless mode, enabling the agent to auto-continue execution until milestones are met.

Installation

To integrate this skill into your environment, use the OpenClaw command-line interface. Run the following command in your terminal: clawhub install openclaw/skills/skills/feiskyer/autonomous-skill Ensure your project root is configured correctly, as the skill creates hidden directories to track task states.

Use Cases

• Complex Software Development: Use it to build an entire REST API from scratch, where individual endpoints need to be designed, coded, and tested across separate sessions.
• Refactoring Projects: Manage large-scale codebase migrations or refactoring efforts that involve touching dozens of files over several days.
• Research and Data Gathering: Orchestrate multi-step data collection tasks that require web searching, file parsing, and summary generation.

Example Prompts

1. "I need to build a secure authentication module for my Express app. This is a long-running task, please start the autonomous workflow."
2. "Autonomous: Refactor the entire directory structure of the frontend components to use the new architecture pattern."
3. "Check the status of my 'build-rest-api' task and continue execution from where it left off."

Tips & Limitations

• Persistence: Always ensure your project root is initialized as a git repository or a stable workspace, as files are created in the .autonomous/ folder.
• Task Naming: The skill automatically truncates names to 30 characters. Keep descriptions concise to avoid unexpected truncation issues.
• Limitations: The skill is not a substitute for human architectural oversight. It works best when the Initializer Agent provides a high-quality task_list.md at the start. Regularly review the progress.md file to ensure the agent is not diverging from your original goals.