Every Thursday at 16:00, the whole cycle read back to you: the issues that shipped, the ones with no owner, the two that are the same crash. You tick the changes you want.
The cycle closes Friday and the board claims 23 issues are In Progress. Six have not moved since the ninth, four have no assignee, and two are the same crash filed twice.
A companion reads all 23 with their comments, checks the pull request behind each one, and comes back with the specific places the board and reality disagree. Nothing moves until you say so.
01
What lands on your desk on Thursday afternoon
A companion pulls the cycle, reads every issue with its comments, then opens the pull request in the tab you are already signed into. It does that for all 23 without skipping the boring ones.
Each disagreement arrives as a proposal you tick through rather than an edit that already happened:
“ENG-412 has a merged PR from Tuesday and no comment since. Move to Done.”
“ENG-388 has no assignee and nine days of silence. Give it an owner or take it out of the cycle.”
“ENG-401 and ENG-457 describe the same stack trace. Make one a sub-issue of the other.”
02
How do you turn a support thread into a real issue?
Point a companion at the thread.
It reads the customer email, the screenshot pasted into a channel and the four messages where the repro steps got split, then writes the issue: a title naming what breaks, the steps in order, a label it looked up rather than invented, a priority and an owner.
It attaches the source as a titled link, so the ticket keeps a route back to the email instead of stopping at “reported by support”. Turning scattered evidence into records is the same motion as data extraction, pointed at your tracker.
03
What can it set, and what stays manual?
The surface is narrower than Linear itself, deliberately.
Sets: state, priority, assignee, labels, project, parent issue and cycle, plus comments and a titled attachment link.
Lists first, so “move it to Todo” resolves to a real ID before a write: teams, users, projects, cycles, labels and workflow states.
Estimates stay with you. No story points, so estimate-driven planning stays a human meeting.
Archive is the strongest action available, and Linear can restore it.
04
Run the same grooming pass every week
Describe the sweep once and save it as a routine: stale In Progress work, unassigned issues, issues whose linked PR merged days ago. It runs the day before your cycle closes and reports. Operations covers how to scope one so it reports rather than rearranges.
Reconciling the board against merged pull requests, a support inbox and a spreadsheet of promises made on calls is four tabs and a judgement call. A companion holds all four open at once and writes down what it found.
Yes. Linear is a native integration you connect once. A companion searches and reads issues, creates them, updates state, priority, assignee, labels, project, parent and cycle, comments, attaches a link, and archives. Changes to shared issues are staged for your approval.
That is the strongest case for it. Point a companion at the email, the support conversation or the channel where the report arrived. It writes the title and the reproduction steps, picks from your existing labels, and shows you the draft before it creates anything.
Describe the sweep once, then save it as a routine that runs the day before your cycle closes. It reports and proposes; you approve the moves.
Consequential actions wait for your approval, so nothing moves on a board your team reads without a person saying yes. Archive is a soft delete Linear can restore, and there is no hard-delete tool at all.
Both are native integrations, so one companion reads a merged pull request and updates the issue it closes without you copying identifiers between tabs. Download Strawberry free and see getting started.
Linear Strawberry