Back to Skills

qa

交互式 QA 会话,用户以对话方式报告缺陷或问题,代理创建 GitHub issues。后台探索代码库以获取上下文和领域语言。适用于用户想报告缺陷、执行 QA、以对话方式提交 issues,或提到 “QA session” 时。

2,530stars207forksUpdated 7/28/2026

Security Assessment

Safe(94/100)
Security Score94/100

About qa

qa is a Chinese-language (translated) skill from the mattpocock-skills-zh-cn collection (deprecated path) that runs an interactive QA session: the user reports bugs or problems conversationally, and the agent clarifies, explores the codebase for context, and files durable, user-focused GitHub issues written in the project's domain language. The problem it solves is turning informal bug reports into well-scoped, long-lived issues that stay meaningful even after major refactors.

For each problem raised, the workflow is: listen and ask only 2-3 short clarifying questions (expected vs. actual, steps to reproduce, stable vs. intermittent); in the background launch an Explore sub-agent to learn the area's domain language (checking UBIQUITOUS_LANGUAGE.md), understand intended behavior, and find the user-facing boundary—without hunting for a fix; assess whether it is a single issue or should be broken into several independent, parallelizable issues; and then file with gh issue create using provided templates (What happened / What I expected / Steps to reproduce / Additional context, plus a breakdown template with Parent issue and Blocked-by fields). Issue bodies deliberately avoid file paths and line numbers, describe behaviors rather than code, require reproduction steps, and stay concise. It prints the resulting URLs, summarizes blocking relationships, and continues until the user is done.

It targets developers and QA practitioners who want to convert conversational bug reports into clean, durable GitHub issues. Its actions are read-only codebase exploration plus issue creation via the official gh CLI—there is no destructive or credential-handling behavior.

FAQ

What does a QA session produce?

Durable, user-focused GitHub issues filed via gh issue create, using structured templates and the project's domain language, with URLs printed and blocking relationships summarized.

Does it try to fix the bugs?

No. It launches an Explore sub-agent only to gather context and domain language, not to find or apply fixes; issues describe user-facing behavior, not internal code.

What are the prerequisites?

The gh CLI authenticated for issue creation and a codebase to explore. It optionally reads UBIQUITOUS_LANGUAGE.md to adopt the project's terminology.

Why do issue bodies avoid file paths and line numbers?

So issues stay durable and meaningful after major refactors—they describe behaviors rather than code that will go stale.

Can it split one report into multiple issues?

Yes. When a report spans independent areas or distinct failure modes, it creates many thin issues in dependency order, marking blocking relationships honestly to maximize parallelism.

All Files

2 files
agents/openai.yaml0.1 KB
View
SKILL.md3.1 KB
View

Install qa

Download and extract the skill files to your .claude/skills/ directory.

Quick Setup:

  1. Copy the skill folder to .claude/skills/
  2. Claude will automatically detect and use the skill