Reproduce repository failures and generate minimal verified git patches.
npx skills add https://github.com/curated-skills/LinguaClaw --skill swe-agent
Please help me install this Agent Skill. Skill: swe-agent Source: https://github.com/curated-skills/LinguaClaw/tree/main/harnesses/artifacts/swe-agent Command: npx skills add https://github.com/curated-skills/LinguaClaw --skill swe-agent
Resolving repository issues reliably by reproducing the failure, applying the smallest credible fix, and proving the fix with targeted verification.
Ask the AI to run the swe-agent workflow: reproduce the issue with a file-backed script, implement the minimal fix using the provided editor workflow, then verify using the reproduction path and the relevant tests before handing off the final git patch.
High-intent search queries and answers about installing and using this skill.
A reproduction-first workflow creates a concrete failing hypothesis and a rerunnable script from your issue description. It then applies the smallest non-test code change set that addresses the root cause while preserving test integrity.
Verification with traceability runs the relevant reproduction and test commands, recording what succeeded or failed. It includes final reflection that highlights risks and remaining uncertainty to guide any necessary retries.
Yes, this workflow applies to bug-fix scenarios where targeted edits, rerunnable reproduction scripts, and command-based verification are required across codebases. It uses a structured editor surface for focused iteration with recorded commands and output.
Your reproduction script might fail to generate a minimal git patch if the reproduction-to-patch loop lacks a structured editor surface or command recording. Successful debugging requires creating a concrete failing hypothesis and applying the smallest non-test change set to address the root cause.
Yes, you need a reproduction-to-patch loop to ship minimal verified patches for repository issues. This loop uses a structured editor surface for focused iteration, rerunnable scripts, command-based verification, and final reflection on risks and remaining uncertainty.
daodaoeduCreate a reproducible GitHub issue with reproduction steps and environment details.
juicesharpOrchestrate multi-agent code review workflows to generate structured review documents from patch diffs.
tesserineCraft work-unit tracker records with observable outcomes and acceptance criteria.
Chang-Shih-YungAggregate type checker, linter, and test results into a 0-10 composite score.
