Mini Swe Agent

What This Skill Does

The Mini Swe Agent is an autonomous software engineering assistant designed to handle complex coding tasks within your repository. Acting as an intelligent sub-contractor, this agent leverages the mini-swe-agent CLI to perform end-to-end codebase exploration, analysis, and modification. It is specifically built for scenarios that require deep contextual understanding of file relationships and complex logic implementation, automating the iterative process of coding, testing, and refining.

Installation

To integrate the Mini Swe Agent into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/c1nderscript/mini-swe-agent

Alternatively, you can manually bootstrap the skill by creating the directory at ~/.openclaw/skills/mini-swe-agent and populating the SKILL.md file with the configuration provided by the author. Ensure your environment has the necessary dependencies to run the mini CLI command.

Use Cases

The skill is best utilized for multi-file refactoring, implementing new features that require touching multiple parts of the codebase, or investigating non-trivial bug reports. If you find yourself needing to search for definitions, trace variable usage across modules, or write tests for complex features, this agent is your primary tool.

Example Prompts

1. "Mini Swe Agent, please refactor the user signup flow to include an email verification step, creating any necessary helper functions in the process."
2. "I am seeing intermittent 500 errors in the payment processing module. Use the mini-swe-agent to investigate the logs and fix the underlying race condition."
3. "Implement a new GitHub issue resolution: Add a markdown exporter for the current project documentation structure."

Tips & Limitations

To get the most out of this agent, always use the --yolo flag to prevent unnecessary interruptions. Remember that the agent is designed for complex tasks; avoid using it for simple one-line changes or minor typos, as it may be overkill and consume more compute resources than necessary. Always verify the output of the agent, especially when it modifies core logic or security-sensitive code. By keeping the problem statement concise and descriptive, you allow the agent to better understand the scope and intent of the requested changes.