🔍 PR Sentinel

The QA engineer's eye on every pull request — powered by Claude AI.

· Open source · MIT License

---

The problem

Every team now has an AI that reviews code quality.
Nobody has an AI that thinks like a QA engineer.

A code reviewer asks: "Is this well written?"
A QA engineer asks: "Does this do what the ticket said? What could break in production?"

PR Sentinel adds the QA perspective to every PR — automatically, before QA is ever pinged.

---

What it does

When a PR is opened or updated, PR Sentinel:

1. Reads the diff and the PR description / acceptance criteria
2. Sends it to Claude with a QA-focused prompt
3. Posts a structured QA analysis as a PR comment

The comment includes:

• AC Coverage — does the code match the acceptance criteria?
• Risk level — LOW / MEDIUM / HIGH
• Functional gaps — what could break that the developer didn't test
• Suggested test scenarios — concrete Given/When/Then cases
• Regression risk — which existing flows are touched
• Open questions — what QA needs answered before sign-off

---

Setup — 3 steps

1. Add your Anthropic API key to GitHub Secrets

In your repo: Settings → Secrets and variables → Actions → New secret

Name: ANTHROPIC_API_KEY
Value: your key from console.anthropic.com

2. Copy the workflow and prompt into your repo

# From your repo root
mkdir -p .github/workflows prompts

curl -o .github/workflows/pr-sentinel.yml \
  https://raw.githubusercontent.com/doqaland/pr-sentinel/main/.github/workflows/pr-sentinel.yml

curl -o prompts/qa-review.md \
  https://raw.githubusercontent.com/doqaland/pr-sentinel/main/prompts/qa-review.md

3. Open a PR

PR Sentinel runs automatically. Look for the comment within ~60 seconds.

---

Customization

The prompt file prompts/qa-review.md is the core of PR Sentinel.
Edit it to add domain context, risk areas, or your team's test conventions.

See docs/customization.md for examples.

---

Example output

See examples/sample-output-feature.md
and examples/sample-output-bugfix.md

---

Why Claude?

Claude's extended reasoning handles long diffs better than most models, and its instruction-following makes the structured output format reliable. You bring your own API key — no data leaves your control.

---

License

MIT — use it, fork it, adapt it for your team.

---

Built by DoQALand — QA automation for teams who ship fast.