Linear Can Draft a QA Plan When a PR Is Ready for Testing. Treat It as a First Draft

Linear Loops can now start from a GitHub label and draft a QA plan. Here's what that plan will catch, what it will miss, and how to set it up so testers still own the risk.
Quick answer
Since October 22, Linear Loops can be triggered by GitHub and Slack events, including drafting a QA plan when a pull request gets a 'ready for testing' label. The draft is good at listing touched areas and obvious edge cases, but it can't see config, data, past incidents or customer context, so teams should use a fixed template and keep a human-written risk section.
Key takeaways
- Linear Loops now trigger from GitHub and Slack, and can be saved as templates with set reasoning levels.
- A drafted plan lists what's in the diff. The risks that break releases often aren't in the diff.
- Use a fixed plan template and reserve a 'known risks' section for people.
- Define what the 'ready for testing' label means before automating on it.
Linear shipped an update on October 22 that will land on a lot of QA desks without much warning. Its Loops feature can now start from GitHub and Slack events, and one of the examples Linear gives is to draft a QA plan when a pull request gets a "ready for testing" label.
So the label goes on, an agent reads the PR, and a test plan shows up before you've opened the ticket.
I think that's useful. I also think it's easy to misuse. A drafted plan is a list of things to check. A QA plan is a decision about where the risk is. Those aren't the same, and the gap between them is your job.
What changed in Linear
From the changelog:
- Loops are Linear's automations that respond to changes in Linear and connected tools. They can now be triggered by events in Slack and GitHub.
- You can save a loop as a template so other teams in your workspace can reuse it. Linear's own example is sharing a bug-triage loop.
- Each loop has a reasoning level: Auto, which follows your workspace default, or Low, Medium or High. Linear pitches it as a way to balance speed, cost and how hard the task is.
The "ready for testing" example is the one that matters for QA. It means the trigger for test planning can move from "a tester picked this up" to "a label was applied."
What a drafted plan will be good at
An agent looking at a PR can see the diff, the title, the description and probably the linked issue. From that it can do a decent job of:
- Listing the screens and endpoints the change touches.
- Turning acceptance criteria into checks.
- Suggesting the obvious edge cases: empty states, long strings, bad input, permissions.
- Reminding you about regression areas near the changed files.
That's the boring half of test planning, and it's fine to hand it off. If you've ever stared at a PR at 6pm trying to remember what else calls that function, a first draft helps.
What it won't see
The things that sink releases usually aren't in the diff.
- Config and data. A feature flag that's on in staging and off in production. A migration that runs fine on a tidy test database and slowly on a real one.
- What broke last time. Your team knows the checkout page falls over when the tax service is slow. The PR doesn't say that.
- Who uses it and how. The one big customer who exports 50,000 rows every Monday. The support ticket from last week that sounds a lot like this change.
- What the PR doesn't touch but should. A missing change is invisible in a diff. The agent can only read what's there.
A plan written from the PR alone will be confident and tidy and still miss the thing that pages someone on Saturday.
How I'd set it up
If your team turns this on, a few rules keep it honest.
1. Write the template before you write the loop. Decide what a QA plan looks like on your team. For example: scope, risk areas, test data needed, environments, what's automated, what's manual, and what's explicitly out of scope. Ask the loop to fill that structure. A free-form plan is harder to review and easier to skim past.
2. Leave one section for a human. Make "Known risks from the team" a required section that the agent can't fill. A tester or the developer adds two or three lines. That's where the flag, the slow query and the angry customer go.
3. Don't let the plan replace the conversation. A drafted plan can make it feel like testing has started when nobody has thought about it yet. Keep the quick chat with the developer. Five minutes of "what are you worried about?" still finds more than any template.
4. Pick the reasoning level on purpose. Linear lets you set it per loop. Test planning is one place where you probably want more reasoning rather than less, because a shallow plan costs you more in missed bugs than it saves in speed. Try it on a few real PRs at Medium and High and compare the output before you settle.
5. Share it as a template once it's good. Templates are the other half of this release. Once one team has a QA loop that people actually read, save it so other teams start from that instead of from scratch.
Watch for the label getting cheap
Here's a quieter risk. If adding "ready for testing" now produces a neat plan automatically, people may start adding the label earlier, before the work is really ready. You'll get plans for half-finished branches and a pile of noise.
Agree on what the label means before you wire a loop to it. A short definition works: the build passes, it's deployed to a test environment, and the description says what changed and why. If those aren't true, the label stays off.
How to tell if it's working
After two or three weeks, look at a handful of shipped features and ask:
- Did the drafted plan cover the bugs we actually found?
- Which bugs did it miss, and were those in the "known risks" section a human added?
- Are testers editing the plans, or just scrolling past them?
If people are scrolling past, the plan is too long or too generic. Cut the template down. If the misses all sit in areas the human section covered, good. That's the split working as designed.
Who does the thinking
A test plan was never valuable because it was written down. It was valuable because someone thought hard about what could go wrong. Let the loop do the listing. Keep the thinking with the people who know where your product breaks.
Frequently asked questions
Loops can now be triggered by Slack and GitHub events, saved as workspace templates, and set to Auto, Low, Medium or High reasoning.


