live-swe-agent

Iteratively debug, patch, and verify software issues within a repository.

43|2|Updated Mar 26, 2026
One-click install
npx skills add https://github.com/curated-skills/LinguaClaw --skill live-swe-agent
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: live-swe-agent
Source: https://github.com/curated-skills/LinguaClaw/tree/main/harnesses/artifacts/live-swe-agent
Command: npx skills add https://github.com/curated-skills/LinguaClaw --skill live-swe-agent

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

It solves the problem of getting stuck when a software engineering issue requires multiple measurement-and-repair iterations to converge on a stable fix.

Core Features & Use Cases

  • Guided iterative workflow: Maintains a disciplined loop of understanding, synthesizing helpers, reproducing, fixing, verifying, and reflecting within the same repo.
  • Command-first, tool-building approach: Uses shell execution for next steps while creating small, file-backed helper scripts to make future actions more precise and rerunnable.
  • Focused, guardrailed edits: Keeps changes within the repository and concentrates edits in source files, avoiding configuration/test drift unless explicitly justified.

Quick Start

Have an AI agent run live-swe-agent on your repository issue description, iterating with single-step terminal commands until the reproducer passes and the fix is ready for handoff.

Frequently Asked Questions about live-swe-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate debugging and patching when fixing a software issue requires multiple iterations?

Automated iterative debugging requires a disciplined loop of observing failures, synthesizing helper scripts, and refining source edits. This approach executes targeted bash commands within a repository, converging on a stable fix by repeatedly measuring and repairing the issue until verification passes.

What is a reproducer-first workflow for repository changes and when do I need it?

A reproducer-first workflow creates a file-backed script to reliably trigger the software issue before attempting any fixes. You need this process for complex debugging or patching tasks to ensure that source edits are validated against a consistent failure state throughout the iteration loop.

How can I synthesize helper scripts to improve bash automation during software verification?

Synthesizing helper scripts involves creating small, file-backed terminal commands during the debugging workflow. This makes future actions more precise and rerunnable by encapsulating complex verification steps into executable scripts, allowing the agent to focus on refining the actual source code edits.

Can I use an agentic workflow to keep repository changes focused on source files and avoid test drift?

Yes, an agentic workflow can guardrail repository changes by concentrating edits in source files and avoiding configuration or test drift. It justifies any non-source modifications explicitly, ensuring that the iterative debugging and patching process maintains strict boundaries within the repository.

What's the best way to converge on a stable fix when repeatedly observing software failures?

The best way to converge on a stable fix is through a self-reflection step that decides whether to synthesize new tools for future iterations. This iterative software engineering process repeatedly observes failures, refines the approach, and executes targeted commands until the reproducer passes.

Why does iteration get stuck when debugging complex software issues and how does self-reflection help?

Iteration gets stuck when a software issue requires multiple measurement-and-repair cycles without a clear path to convergence. Self-reflection helps by analyzing the failed attempts and deciding whether to synthesize new helper tools, enabling the workflow to evolve and break through the debugging plateau.