Dev Flow

How it works

The mechanics of this kit, plainly — what runs and when, where the data lands, and what you look at.

• Dev digest runs weekday mornings · 8am and keeps the Dev Flow dashboard fresh. Weekday mornings — whose move it is, and what has gone stale.
• What is captured lands in the PR log dataset, which lives at Data/PR Log in your vault — yours to open, search, and edit like any other file.
• What is captured lands in the Dev log dataset, which lives at Data/Dev Log in your vault — yours to open, search, and edit like any other file.
• What is captured lands in the Debt list dataset, which lives at Data/Debt List in your vault — yours to open, search, and edit like any other file.
• The Dev Flow dashboard reads from the dataset and shows, at a glance: prs on the log · merged this week · issues on you · reviews you owe · whose move · by repo · go through the debt list · the debt list, by where it stands · shipped and broke, lately · how this board works.

The wiring, in one breath: Dev digest writes to PR log · Dev digest writes to Dev log · Incident capture writes to Dev log · PR log shown by Dev Flow · Dev log shown by Dev Flow · Dev Flow kept fresh by Dev digest · Incident capture given to Relay · Dev lookback given to Relay · Debt list keeper writes to Debt list · Debt list shown by Dev Flow · Write-ups reads Dev log · Queue reader given to Relay · Write-ups given to Relay · Debt list keeper given to Relay.

Getting the most from it

A kit earns its place when it changes what you look at and what you say, not when it is installed. This is the rhythm that makes it pay.

• Open Dev Flow before you open a tab. The digest has already sorted every open pull request by whose move it is and how long it has sat, and counted the issues assigned to you.
• Before you review, ask Relay what a pull request does. Queue reader says what it claims, what it touches and how big it is, from the pull request itself; the review, and any comment on it, stay yours.
• Say incidents, deploys and decisions in passing — "prod was down twenty minutes, bad migration", "api 2.14 went to production", "we decided to drop the legacy endpoint" — and Incident capture puts them on the record. When an incident is over, ask Write-ups for a blameless write-up.
• Ask Write-ups for your stand-up note in the morning, or for the release notes of what merged since the last deploy. Both come from the record and what you tell it, never from a guess.
• Name technical debt as you trip over it and Debt list keeper keeps it in your words. Now and then, press Go through the debt list on the board and rate each item again.
• Ask Dev lookback when the argument starts: "what shipped in June", "why did we pick that database". Tell Relay which repositories matter and what counts as stale, and remember there is no view of CI here: the board is pull-request state and nothing more.

Make it yours

Nothing here is fixed. The pieces are ordinary agents, routines, datasets and dashboards, so anything you know about shaping those applies. Retire whatever earns nothing.

Things people usually change first

Say any of these in chat, or edit them on the kit’s page in the app.

• How many days count as stale, separately for reviews you owe and your own waiting work.
• Which repositories it watches.
• When the digest arrives, or turning the cards off and just keeping the board.
• What is worth putting on the dev log — some people log only incidents, some log every decision.
• How your stand-up note is laid out, and which repositories the release notes cover.
• What counts as debt worth keeping, and how you rate it.

Your first run

Installing starts a short setup conversation — not a form. Answer roughly, correct later.

1. Adding the Kit offers the first digest right away: once GitHub is connected on the Connectors screen, start it with one tap and Relay reads your repositories — every open pull request you wrote or were asked to review, whose move it is and how long it has waited. When it finishes, Relay leaves a short note and invites two minutes together on which repositories matter and what you count as stale. Rather start without it? Tell Relay something that shipped or broke, and it goes on the dev log.
2. It tells you only what has genuinely gone stale. A queue in good shape produces no card at all.
3. Mention an incident, a deploy or a decision in passing and it lands on the dev log, in the words you used at the time.
4. It cannot see build status — there is no CI tool on the connector — so it never reports whether a build passed or main is green.

Watch the first pass, then tell it what you would have done differently — that correction teaches it more than any setup could.

The essentials

Learning, preferences and run costs work the same in every kit — so they are covered once, in the Kits guide.

Living with a Kit — the Kits guide →

← All kits