next-qa

Runs one unattended QA iteration that builds a queued test task and opens a PR on auto-qa.

5.4k|571|Updated Nov 1, 2025
One-click install
npx skills add https://github.com/breaking-brake/cc-wf-studio --skill next-qa
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: next-qa
Source: https://github.com/breaking-brake/cc-wf-studio/tree/main/.claude/skills/next-qa
Command: npx skills add https://github.com/breaking-brake/cc-wf-studio --skill next-qa

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Feature branches often merge with only build and type checks as gates, so behavioral regressions slip through undetected. This Skill closes that gap by autonomously building an automated test suite, one queued QA issue per invocation, without ever touching product source code.

Core Features & Use Cases

  • Serialized QA iterations: Stewards any in-flight QA PR (merging on green CI, fixing red builds) before picking exactly one queued qa issue per run.
  • Scope-safe test authoring: Creates only test files, test configuration, and CI workflow updates on the auto-qa integration branch; product bugs are filed as bug issues with skipped regression tests instead of being fixed in place.
  • Full PR lifecycle automation: Syncs auto-qa with main, branches, runs pnpm build && pnpm check && pnpm test, logs to docs/qa-log.md, opens a PR, and squash-merges only on green CI.
  • Use Case: Say "next qa" and the agent picks the highest-value queued test task, implements a deterministic regression test for a known bug, and merges it into auto-qa after CI passes.

Quick Start

Ask the agent to run the next QA iteration by saying "next qa" so it stewards any open QA PR and builds one queued test issue on auto-qa.

Frequently Asked Questions about next-qa

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

FAQPage Schema
How do I run an autonomous QA iteration on my repository?

Invoke the skill by saying "next qa" or "QAタスク". It checks for an in-flight QA PR first, then selects one queued issue labeled qa, implements the test on a branch off auto-qa, and opens a PR that squash-merges on green CI.

How does the skill choose which test to write next?

It picks the eligible qa issue with the best protection-to-effort ratio, preferring test infrastructure, regression tests for real bugs, then unit tests for pure logic in packages/core, cli, and mcp. Every task must protect a user-facing behavior, be deterministic, and ship in one PR.

Can the QA loop fix product bugs it discovers?

No. The loop never edits packages/*/src. When a test surfaces a real bug, it files a bug issue and lands the test in a skipped state referencing that issue; the feature loop on auto-dev performs the actual fix.

What happens when CI fails on a QA pull request?

The skill fixes the failure and re-pushes, up to three attempts. After three failed attempts it leaves the PR open, labels it needs-attention, marks the qa-log entry as blocked, and stops rather than forcing a merge.

Why does the skill use a separate auto-qa branch instead of main?

The auto-qa integration branch is a sibling of auto-dev, both branching from main and promoted by a human. Agent merges are allowed only into auto-qa via PR with green CI, keeping autonomous test work isolated from production until human promotion.