qa-agent

Independently verifies GitHub issues labeled flow:ready-qa with proof in the running app.

96|11|Updated Feb 8, 2026
One-click install
npx skills add https://github.com/thefrederiksen/devthrottle --skill qa-agent-thefrederiksen
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: qa-agent
Source: https://github.com/thefrederiksen/devthrottle/tree/main/.claude/skills/qa-agent
Command: npx skills add https://github.com/thefrederiksen/devthrottle --skill qa-agent-thefrederiksen

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Development agents can report work as done without real verification, leaving broken features merged and messy repositories behind. This Skill acts as an independent QA gate that re-tests every acceptance criterion itself, with screenshots and API evidence, before an issue is closed. ## Core Features & Use Cases - Independent verification loop: Pulls the oldest GitHub issue labeled flow:ready-qa, rebuilds and runs the change in an isolated session slot, and judges each acceptance criterion Expected vs Actual with its own screenshots. - Pass, fail, or park outcomes: Passes merge the PR via squash and close the issue (flow:done), failures bounce back to the developer with a reproducible written defect (flow:qa-failed), and unresolvable conflicts escalate as a parked flow:needs-human PR. - Zero-mess cleanup gate: Enforces an empty git status, no stashes, no orphaned branches, and no dangling PRs before reporting any result. - Use Case: A team running an autonomous implementation loop needs each completed issue verified by a separate identity before merge; the QA Agent processes the ready-qa queue end to end, producing committed HTML proof reports under docs/cencon/proof/. ## Quick Start Ask the agent to run the QA queue and verify the next flow:ready-qa issue in the repository.

Frequently Asked Questions about qa-agent

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

FAQPage Schema
How do I run the QA verification queue on GitHub issues?▼

Trigger the QA Agent with a phrase like "run the QA queue" and it picks the oldest open issue labeled flow:ready-qa via the gh CLI. It verifies each acceptance criterion independently, then labels the issue flow:done or flow:qa-failed and loops until the queue is empty.

How does the QA Agent verify acceptance criteria independently?▼

It checks out the PR branch, allocates its own isolated session slot, builds the change itself, and launches the app via a per-slot scheduled task. Each criterion is reproduced through the UI or Control API with a screenshot judged Expected vs Actual, never trusting the developer's report.

Can the QA Agent merge pull requests to main automatically?▼

Only when driven by the implementation-loop: a clean pass squash-merges the PR and deletes the branch, after verifying the merged result builds with zero errors. A standalone QA session stops at flow:done and leaves merging to a human, and conflicts escalate to flow:needs-human.

Does the QA Agent work with Azure DevOps work items?▼

Yes, in devops mode the Azure DevOps work item is the tracker: its description provides the acceptance criteria and verdict comments are posted via az boards work-item update. The GitHub PR, proof directory, and cleanup gates remain unchanged.

What happens when a QA verification fails?▼

The agent writes a specific reproducible defect with exact steps, Expected vs Actual, and failure screenshots committed under docs/cencon/proof/issue-<n>/. It comments on the issue, labels it flow:qa-failed, and hands it back to the Developer Agent without fixing code itself.

Why does the QA Agent forbid git stash during cleanup?▼

A stash leaves the working tree empty while hiding work in progress, which fakes a clean repository state. The skill requires all WIP to be committed to the PR branch instead, and every cleanup gate asserts git stash list contains nothing the agent created.