惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Stack Overflow Blog
Stack Overflow Blog
量子位
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
美团技术团队
小众软件
小众软件
aimingoo的专栏
aimingoo的专栏
Recent Announcements
Recent Announcements
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
Microsoft Security Blog
Microsoft Security Blog
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
酷 壳 – CoolShell
酷 壳 – CoolShell
J
Java Code Geeks
V
V2EX
大猫的无限游戏
大猫的无限游戏
D
DataBreaches.Net
博客园 - Franky
爱范儿
爱范儿
T
Tailwind CSS Blog
A
About on SuperTechFans
Google DeepMind News
Google DeepMind News
博客园_首页
B
Blog RSS Feed
博客园 - 司徒正美
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
Your Kanban board is lying to you (and Git knows it)
Matías Denda · 2026-04-29 · via DEV Community

Look at your team's board right now.

How many tickets are in "In Review" that haven't been looked at in three days? How many are in "In QA" even though nobody's tested them? How many jumped straight from "In Progress" to "Done" without ever appearing in the intermediate columns?

If you're like most teams, the answer is "a lot." Your board is not reflecting reality. It's a fiction your team politely maintains.

Here's the uncomfortable truth: your Git repository knows the real state of every ticket. Your board just hasn't caught up.

This article is about bridging that gap.

The problem: columns designed by process people, not by engineers

Most boards are designed by someone who thinks about process, not about what actually happens in Git. That's why you end up with:

  • "Ready for Development" — a holding pen for tickets nobody wants to admit aren't ready. If it's prioritized and has detail, it's "To Do." If it doesn't, it's still "Backlog."
  • "Ready for QA" — this implies a manual handoff. If the code is merged and deployed to a place QA can access, it's in QA. The deployment is the handoff, not a ticket drag.
  • "Deployed" as a column separate from "Done" — if the code is in production and verified, it's done. If it's deployed but not verified, it's still in QA (in production). A column between them means nobody is verifying production deployments. These columns feel orderly on a wiki diagram. In practice, they're fiction. Tickets skip them. Developers lie about their state. The board drifts from reality.

The test: can someone verify the column from Git alone?

Here's the rule that cuts through all the noise:

Every column must correspond to a state that someone can verify, independently of anyone else's memory.

"In Progress" passes this test — is there a branch with commits? Verifiable.

"Code Review" passes — is there an open PR? Does it have an approving review? Verifiable via gh pr view or the hosting platform's API.

"Ready for Testing" fails — ready according to whom? Tested where? This is a human statement, not a Git state.

If a column can't be verified from the repo, one of two things is happening:

  1. The column represents a task-flow state, not a code-flow state. That's fine — but it needs to be a lateral exit, not part of the main flow. "Blocked" and "Needs Spec" are valid columns, but they're paused states, not progression states.
  2. The column is fiction. Remove it, or fix the process so it becomes real. ## Two dimensions that most boards collapse into one

The biggest insight I had while designing boards with teams: there are two independent dimensions of ticket state, and teams keep smashing them into a single horizontal flow.

Code-flow columns (left to right): To Do → In Progress → In Review → In QA → Done. These map directly to Git states.

Task-flow columns (lateral exits): Blocked, On Hold, Needs Spec, Waiting for Client. These don't map to Git at all — they reflect the state of the work, not the state of the code.

A ticket in "Blocked" still has a Git branch open. It hasn't moved backward in the flow — it's paused. When you separate these two dimensions, you prevent the common mistake of adding "Blocked" as a column between "In Progress" and "In Review," which breaks the flow and confuses everyone.

Diagram showing two independent dimensions of a board. The top row shows code-flow columns arranged horizontally with arrows: To Do, In Progress, In Review, In QA, Done — each labeled with its Git equivalent. The bottom row shows task-flow columns arranged laterally: Blocked, On Hold, Needs Spec, Waiting for Client — with dashed borders indicating they are lateral exits. Dashed arrows descend from code-flow states to task-flow states, showing that any code-flow state can derive to a lateral exit at any time.

The reality: columns that get skipped

On paper, tickets move through every column. In practice, developers skip half of them.

Tickets jump from "In Progress" directly to "Done" — skipping "In Review" and "In QA" entirely. Why? Because the developer merged the PR, verified it themselves in production, and dragged the ticket in one motion.

This isn't laziness. It's the board not matching reality. If the team doesn't have a dedicated QA person or a staging environment, the "In QA" column is fiction. Tickets skip it because the state it represents doesn't exist.

The rule: if a column is consistently skipped, remove it or fix the process. Either add the missing step (a real QA phase, a real staging environment) or remove the column that pretends it exists. A board with skipped columns is a board that lies.

The solution: let Git move the tickets

The best boards are updated by Git events, not by humans dragging cards. Most tracking systems support this natively:

  • Opening a PR moves the ticket to "In Review" — automatically
  • Review approval can move it to "Approved" — automatically
  • Merging the PR moves it to "In QA" or "Done" — automatically
  • Tagging a release can move it to "Done" — automatically Here's a GitHub Actions workflow that moves a GitHub Projects card when a PR is opened or merged:
# .github/workflows/board-automation.yml
name: Board automation
on:
  pull_request:
    types: [opened, closed]

jobs:
  move-card:
    runs-on: ubuntu-latest
    steps:
      - name: Move to In Review when PR opens
        if: github.event.action == 'opened'
        uses: actions/github-script@v7
        with:
          script: |
            // Move the linked issue to "In Review" column
            // Uses GitHub Projects V2 API
            const query = `mutation($projectId: ID!, $itemId: ID!, $fieldId: ID!, $value: String!) {
              updateProjectV2ItemFieldValue(input: {
                projectId: $projectId, itemId: $itemId,
                fieldId: $fieldId,
                value: { singleSelectOptionId: $value }
              }) { projectV2Item { id } }
            }`;
            // projectId, fieldId, and optionId come from your project's
            // settings — query them once with the GraphQL explorer

      - name: Move to In QA when PR merges
        if: github.event.action == 'closed' && github.event.pull_request.merged
        uses: actions/github-script@v7
        with:
          script: |
            // Same mutation, different singleSelectOptionId
            // pointing to the "In QA" column

Enter fullscreen mode Exit fullscreen mode

For Jira, it's even simpler — built-in automation rules handle this:

Rule 1: Trigger: Pull request created. Condition: Issue status is 'In Progress'. Action: Transition to 'In Review'.

Rule 2: Trigger: Pull request merged. Condition: Issue status is 'In Review'. Action: Transition to 'In QA'.

Linear, Shortcut, and Azure DevOps have similar integrations. The specifics differ, but the pattern is universal: Git event → webhook → board transition.

What you need to make this work

The automation is trivial. The discipline it requires is the hard part:

1. Branch names must include ticket IDs.

# Good — automation can match on this
git checkout -b feature/PROJ-247-search-filters

# Bad — automation has nothing to match
git checkout -b sam/search-thing

Enter fullscreen mode Exit fullscreen mode

2. PR descriptions must reference the ticket.

gh pr create --title "feat(search): add filters" \
             --body "Closes PROJ-247"

Enter fullscreen mode Exit fullscreen mode

3. Accept that some transitions stay manual.

"In Progress" is manual — and that's fine. When a developer picks up a ticket, they drag it and create a local branch. Nobody else knows about the branch until they push. Moving the ticket is faster than pushing a WIP commit just to trigger automation, and it avoids polluting Git history with empty commits.

The goal isn't 100% automation. The goal is that the board reflects reality, even when nobody has time to update it.

Reading the board from Git (when you don't trust the board)

Once you accept that Git is the source of truth, you can answer board questions without even looking at the board.

"What's actually in progress right now?" Remote branches ahead of main with no open PR:

git fetch --prune
for branch in $(git for-each-ref --format='%(refname:short)' refs/remotes/origin/ \
                | grep -v 'origin/main$' | grep -v 'origin/HEAD'); do
  ahead=$(git rev-list --count origin/main.."$branch" 2>/dev/null)
  [ "$ahead" -gt 0 ] || continue
  has_pr=$(gh pr list --head "${branch#origin/}" --state open --json number --jq 'length')
  [ "$has_pr" = "0" ] && echo "In Progress: $branch ($ahead commits ahead)"
done

Enter fullscreen mode Exit fullscreen mode

"What's in review?" Open PRs — simplest query, most useful:

gh pr list --state open \
  --json number,title,author,reviewDecision,createdAt \
  --jq '.[] | "\(.number)\t\(.author.login)\t\(.reviewDecision // "PENDING")\t\(.title)"'

Enter fullscreen mode Exit fullscreen mode

"What's merged but not deployed?" Commits on main since the last production tag — the invisible column most teams don't track:

# What's on main that hasn't been tagged yet?
git log --oneline $(git describe --tags --abbrev=0)..origin/main

# Extract ticket IDs — these are "done but undelivered"
git log --pretty=%s $(git describe --tags --abbrev=0)..origin/main \
  | grep -oE '[A-Z]+-[0-9]+' | sort -u

Enter fullscreen mode Exit fullscreen mode

"What's actually in production right now?" Tickets referenced by commits reachable from the production tag:

CURRENT_PROD_TAG=$(git describe --tags --abbrev=0 --match 'v*')
git log --pretty=%s ${CURRENT_PROD_TAG} \
  | grep -oE '[A-Z]+-[0-9]+' | sort -u

Enter fullscreen mode Exit fullscreen mode

Run these for a week. Compare against what the board shows. If they disagree, the board is lying — and now you have the data to fix either the board or the process.

The mindset shift

The board isn't the truth. It's a summary of the truth that drifts whenever someone forgets to drag a ticket.

Git is the truth. Every branch, every PR, every merge is a real event with a real timestamp. You can't forget to tell Git — it records everything automatically, as a side effect of doing the work.

Teams that embrace this get three wins:

  1. The board starts matching reality (because Git drives it, not human memory)
  2. "Where is this ticket?" gets one answer, not three
  3. Standup meetings stop being ticket-hunting exercises and start being real conversations about blockers That's what a good board does. That's what yours could be doing, if you let Git update it.

This post is adapted from Git in Depth: From Solo Developer to Engineering Teams, a 658-page book covering Git the way it's actually used in real engineering teams — including a full chapter on aligning board columns with Git states.

Related: What actually happens when you git merge --no-ff.

See all my articles on Git and engineering practice: dev.to/matutetandil.