<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Steve</title>
    <description>The latest articles on DEV Community by Steve (@steve-o).</description>
    <link>https://dev.to/steve-o</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4245816%2F227e1f5e-48a9-481d-997e-1ec11ba5cf69.jpg</url>
      <title>DEV Community: Steve</title>
      <link>https://dev.to/steve-o</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/steve-o"/>
    <language>en</language>
    <item>
      <title>Writing Acceptance Criteria an AI Agent Can Actually Meet</title>
      <dc:creator>Steve</dc:creator>
      <pubDate>Mon, 16 Nov 2026 02:54:08 +0000</pubDate>
      <link>https://dev.to/steve-o/writing-acceptance-criteria-an-ai-agent-can-actually-meet-38a</link>
      <guid>https://dev.to/steve-o/writing-acceptance-criteria-an-ai-agent-can-actually-meet-38a</guid>
      <description>&lt;p&gt;What happens when you give a vague ticket to a person?&lt;/p&gt;

&lt;p&gt;They ask a question. Or they fill the gap with what they know about your team, your customers and that one time it all went wrong.&lt;/p&gt;

&lt;p&gt;Now give the same ticket to an AI agent.&lt;/p&gt;

&lt;p&gt;It doesn't ask. It fills the gap with a plausible guess, finishes, and reports back with total confidence:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Implemented and verified. ✅&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;And you find out what it guessed in review, reading a diff against a definition of done that nobody wrote down. 😅&lt;/p&gt;

&lt;p&gt;Acceptance criteria are where that gets fixed, before the work starts instead of after. With an agent they matter &lt;em&gt;more&lt;/em&gt;, not less. Every gap gets filled, and nobody tells you which ones.&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Full disclosure: I make Arbituro, a work tracker built around criteria like these. Everything in this post works in any tracker, and only the last section is about Arbituro.&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  🤔 Three Questions for Every Criterion
&lt;/h2&gt;

&lt;p&gt;A criterion an agent can meet answers three questions:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;What will be true?&lt;/strong&gt; An outcome someone can observe, not an activity.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;How will we know?&lt;/strong&gt; The evidence: a test, a check's result, a page someone opens, a command's output.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Who decides?&lt;/strong&gt; A named CI check, or a named person. Not "the team". Not "QA". And definitely not nobody.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Can't answer the third one? Then it isn't a criterion yet. It's a wish. 🌠&lt;/p&gt;

&lt;h2&gt;
  
  
  📏 Eight Rules
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Rule #1: Write Outcomes, Not Activities
&lt;/h3&gt;

&lt;p&gt;"Add tests" describes effort. "Signing up twice with the same address is refused" describes a result. An agent can always do the activity. Only the outcome tells it when to stop.&lt;/p&gt;

&lt;h3&gt;
  
  
  Rule #2: Replace Every Quality Word
&lt;/h3&gt;

&lt;p&gt;Fast, clean, robust, properly, user-friendly. Each one is a gap the agent fills with its own standard, and its standard is... generous. 😇 Swap the word for a number, an example or a named judge.&lt;/p&gt;

&lt;h3&gt;
  
  
  Rule #3: One Thing per Criterion
&lt;/h3&gt;

&lt;p&gt;A criterion with "and" in it can be half met, and then nobody can say met or not met. Split it.&lt;/p&gt;

&lt;h3&gt;
  
  
  Rule #4: Name the Edge Case You Care About
&lt;/h3&gt;

&lt;p&gt;Agents cover the cases you wrote down. So write down the one you'd check first in review.&lt;/p&gt;

&lt;h3&gt;
  
  
  Rule #5: Say What's Out of Scope
&lt;/h3&gt;

&lt;p&gt;Agents love to widen work. A rename here, a little tidy-up there, and suddenly the diff touches forty files. 🧹 One line saying what this job leaves alone keeps the diff reviewable.&lt;/p&gt;

&lt;h3&gt;
  
  
  Rule #6: Let a Check Decide Wherever One Can
&lt;/h3&gt;

&lt;p&gt;A CI check decides "the test passes" on every push, and it never gets tired, bored or hungry. Keep people for judgement: wording, design, whether the change actually solves the problem.&lt;/p&gt;

&lt;h3&gt;
  
  
  Rule #7: Keep Standing Rules Out of Tickets
&lt;/h3&gt;

&lt;p&gt;"Lint passes", "reviewed by a second engineer" and "coverage holds" apply to every change. Enforce them once, in CI and branch protection, so each ticket's criteria say only what's specific to it.&lt;/p&gt;

&lt;h3&gt;
  
  
  Rule #8: Two to Five per Job
&lt;/h3&gt;

&lt;p&gt;Fewer, and something's usually missing. More, and it's probably two jobs in a trench coat. 🧥&lt;/p&gt;

&lt;h2&gt;
  
  
  🔧 Before and After
&lt;/h2&gt;

&lt;p&gt;Enough theory. Let's fix some tickets.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Before:&lt;/strong&gt; "Handle expired reset links properly."&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;After:&lt;/strong&gt; Opening an expired or already used reset link shows "This link has expired" and a button that sends a new one. &lt;em&gt;Decided by the &lt;code&gt;e2e&lt;/code&gt; check.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;"Properly" was carrying that whole ticket. 😂&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Before:&lt;/strong&gt; "Make search faster."&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;After:&lt;/strong&gt; On the seeded test data of 50,000 orders, the first page of search results comes back in under 300 ms at the 95th percentile. &lt;em&gt;Decided by the &lt;code&gt;search-perf&lt;/code&gt; check.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Faster than what? Now we know.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Before:&lt;/strong&gt; "Make checkout mobile-friendly."&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;After:&lt;/strong&gt; At 375 px wide, the checkout form fits without scrolling sideways, and every button is at least 44 px tall. &lt;em&gt;Priya decides, from screenshots at 375 px.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;And Priya knows exactly what she's looking at.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Before:&lt;/strong&gt; "Clean up the billing code."&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;After:&lt;/strong&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Billing behaves exactly as before: every existing billing test passes. &lt;em&gt;Decided by the &lt;code&gt;test&lt;/code&gt; check.&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;InvoiceService&lt;/code&gt; reaches the payment provider only through &lt;code&gt;PaymentClient&lt;/code&gt;. &lt;em&gt;Priya decides, from the diff.&lt;/em&gt;
&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;&lt;em&gt;Not in this job: renaming anything other code calls.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;"Clean up" is how a two-line fix becomes a forty-file diff. That last line is doing real work.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Before:&lt;/strong&gt; "Add tests for the duplicate email bug."&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;After:&lt;/strong&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Signing up with an address that differs from an existing one only in case is refused with "That email is already registered". &lt;em&gt;Decided by the &lt;code&gt;test&lt;/code&gt; check.&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;The new test fails on &lt;code&gt;main&lt;/code&gt;, so it really catches the bug. &lt;em&gt;Sam decides, from the agent's run of the test against &lt;code&gt;main&lt;/code&gt;.&lt;/em&gt;
&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;That second criterion is sneaky, and I love it. More on why in a minute.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Before:&lt;/strong&gt; "Tell customers about the price change."&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;After:&lt;/strong&gt; Customers on the old plan are sent the approved email about the new price. &lt;em&gt;Met when the approved email is sent.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Wait, that one isn't code at all! Right. Agents draft emails and posts too, and done there means the message actually went out. A draft sitting somewhere doesn't count. ✉️&lt;/p&gt;

&lt;h2&gt;
  
  
  ⚖️ Who Decides: a Check, a Person or a Send
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;A CI check&lt;/strong&gt; suits anything a machine can verify. Name it exactly as it appears in the pull request's checks, and watch out for matrix jobs, which get names like &lt;code&gt;test (20)&lt;/code&gt;.&lt;/p&gt;

&lt;p&gt;A check is only as good as the test behind it, though. When the agent writes the test too, the test is what you review. That's why "the new test fails on &lt;code&gt;main&lt;/code&gt;" is up there: it catches a test that passes whatever the code does. Yes, that's a thing that happens. 🙃&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;A person&lt;/strong&gt; suits judgement: wording, design, whether the fix solves the customer's problem. Name one person. When the agent believes the criterion is met, it says what it checked, and that person decides.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;A send&lt;/strong&gt; suits work whose result goes out to someone: a customer email, a release note in a team channel. It's met when the approved message is actually sent.&lt;/p&gt;

&lt;h2&gt;
  
  
  🧾 What a Claim Should Say
&lt;/h2&gt;

&lt;p&gt;When an agent thinks a person-decided criterion is met, the useful report is what it checked, not that it's finished.&lt;/p&gt;

&lt;p&gt;Here's our old friend again:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Implemented and verified. ✅&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;And here's a claim somebody can actually decide on:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Ran &lt;code&gt;npm test -- signup&lt;/code&gt; on main: "refuses an address that differs only in case" fails (expected 409, got 201). On this branch it passes, along with the other 212 tests.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;See the difference? The second one names the command, says what happened on &lt;code&gt;main&lt;/code&gt; and says what happened on the branch. A claim should name the command it ran or the page it opened, and quote what it saw. One with nothing behind it hands the person who decides a decision with nothing to decide on.&lt;/p&gt;

&lt;h2&gt;
  
  
  🤝 Let the Agent Draft, and a Person Agree
&lt;/h2&gt;

&lt;p&gt;Here's the twist: agents are really good at &lt;em&gt;drafting&lt;/em&gt; criteria. Ask one to read the ticket and the code before it starts, and it'll often find the edge case you missed.&lt;/p&gt;

&lt;p&gt;What it shouldn't do is set its own bar. That's like letting me write my own performance review. 😅 The agent proposes, a person accepts or edits, and then the work starts.&lt;/p&gt;

&lt;p&gt;The same goes mid-work. If a criterion turns out to be wrong, the agent should say so and stop, not quietly work to a different one. Changing or dropping a criterion is a person's decision, made on the record with a reason, so the review matches what was agreed.&lt;/p&gt;

&lt;h2&gt;
  
  
  📋 A Template
&lt;/h2&gt;

&lt;p&gt;Copy it, paste it, make it yours:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight markdown"&gt;&lt;code&gt;&lt;span class="gs"&gt;**Done when**&lt;/span&gt;
1. &amp;lt;What will be true&amp;gt;. Decided by the &lt;span class="sb"&gt;`&amp;lt;check&amp;gt;`&lt;/span&gt; check.
2. &amp;lt;What will be true&amp;gt;. &amp;lt;Name&amp;gt; decides, from &amp;lt;evidence&amp;gt;.
3. &amp;lt;What will be true&amp;gt;. Met when the approved &amp;lt;email or post&amp;gt; is sent.

&lt;span class="gs"&gt;**Not in this job:**&lt;/span&gt; &amp;lt;what to leave alone&amp;gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  🏗️ How Arbituro Does It
&lt;/h2&gt;

&lt;p&gt;Okay, the promised product bit. In Arbituro, every acceptance criterion says how it's decided: by a named person, by a CI check, or by an approved message being sent. Agents read the criteria before they start, whether they're Arbituro's own agents or your coding agent connected over MCP.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnodfyxysuhymx8ma9vxn.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnodfyxysuhymx8ma9vxn.webp" alt="Arbituro's agent proposing four acceptance criteria for a new job, each naming the person who decides it, with a way to add your own and a button to agree to all four." width="800" height="958"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Drafting.&lt;/strong&gt; When a job is filed, the project's agent suggests criteria, each naming who decides it, and you accept, edit or decline each one. A coding agent can propose criteria too, naming the check that should decide each, and a person accepts them.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Checks.&lt;/strong&gt; A criterion that names a CI check is decided when that check finishes on the latest commit of a pull request linked to the job: met when it succeeds, not met when it fails.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;People.&lt;/strong&gt; For a criterion a person decides, an agent can only record what it checked. That person is asked, and decides in one press.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Changes.&lt;/strong&gt; Agents never tick, change or remove a criterion. The person who decides one, or the job's owner, can waive it with a reason, so the record shows what changed and why.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  🏁 Wrapping Up
&lt;/h2&gt;

&lt;p&gt;Three questions, eight rules, and somebody (or something) named to decide each criterion. That's really all it takes to stop getting "Implemented and verified. ✅" as a status report.&lt;/p&gt;

&lt;p&gt;Arbituro is free for up to three people at &lt;a href="https://arbituro.com" rel="noopener noreferrer"&gt;arbituro.com&lt;/a&gt;, and the &lt;a href="https://arbituro.com/docs/acceptance-criteria" rel="noopener noreferrer"&gt;acceptance criteria docs&lt;/a&gt; have the details. For the coding side, the other post in this series has Claude Code working a ticket like these over MCP: &lt;a href="https://arbituro.com/blog/claude-code-tickets-over-mcp" rel="noopener noreferrer"&gt;Stop Pasting Tickets into Claude Code&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;What about you? How does your team write acceptance criteria, and have you tried handing them to an agent yet? Bonus points for the vaguest criterion you've ever seen. 😂 Tell me in the comments. 👇&lt;/p&gt;

</description>
      <category>ai</category>
      <category>agile</category>
      <category>productivity</category>
      <category>testing</category>
    </item>
    <item>
      <title>Stop Pasting Tickets into Claude Code: Let It Read Them over MCP</title>
      <dc:creator>Steve</dc:creator>
      <pubDate>Mon, 16 Nov 2026 02:35:14 +0000</pubDate>
      <link>https://dev.to/steve-o/stop-pasting-tickets-into-claude-code-let-it-read-them-over-mcp-if7</link>
      <guid>https://dev.to/steve-o/stop-pasting-tickets-into-claude-code-let-it-read-them-over-mcp-if7</guid>
      <description>&lt;p&gt;How do you hand Claude Code a ticket?&lt;/p&gt;

&lt;p&gt;If you're anything like me, you copy it, paste it into the prompt, add "please" (just in case 😅) and press Enter.&lt;/p&gt;

&lt;p&gt;And it works! Once.&lt;/p&gt;

&lt;p&gt;Then somebody edits the ticket, and the agent never hears about it. Its branch is called &lt;code&gt;fix-stuff&lt;/code&gt;. Nobody can tell which commits belong to which ticket. And the ticket is finished when the agent says so, which is usually the moment it types its favourite sentence:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Implemented and verified. ✅&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;Verified by whom, exactly? 🙃&lt;/p&gt;

&lt;p&gt;So let's build a better loop. It has three parts:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;The agent reads the ticket itself&lt;/strong&gt;, over MCP, so it always works from the current version, including what "done" means.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Its work names the ticket.&lt;/strong&gt; The branch, every commit and the pull request carry the ticket's key, so the ticket shows the work as it lands.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Done is decided outside the agent&lt;/strong&gt;, by a CI check or by a person who can see what the agent actually checked.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;I'll set it up with Claude Code and GitHub. The examples come from Arbituro's demo workspace, where a coffee company is adding a four-pack to its online shop. ☕ Using Jira or Linear? Stick around: a section near the end says what carries over.&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Full disclosure: I make Arbituro, the work tracker in the screenshots. I build it with coding agents too, and its own backlog lives in it, so this is the loop I use myself.&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  🧰 What You'll Need
&lt;/h2&gt;

&lt;p&gt;Nothing exotic:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Claude Code.&lt;/li&gt;
&lt;li&gt;A tracker with a remote MCP server. I'll use &lt;a href="https://arbituro.com" rel="noopener noreferrer"&gt;Arbituro&lt;/a&gt;, which is free for up to three people.&lt;/li&gt;
&lt;li&gt;A GitHub repository with CI. The examples use GitHub Actions.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;There's one bit of setup on the Arbituro side. A workspace owner or admin installs the GitHub app once and links the repository to the project. That's what lets pushes and pull requests find their ticket.&lt;/p&gt;

&lt;h2&gt;
  
  
  🔌 Step 1: Connect the Tracker
&lt;/h2&gt;

&lt;p&gt;One command:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;claude mcp add --transport http arbituro https://app.arbituro.com/mcp
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Then run &lt;code&gt;/mcp&lt;/code&gt; in Claude Code, pick &lt;strong&gt;arbituro&lt;/strong&gt; and sign in. Your browser opens the tracker's sign-in page, and from then on Claude Code works as &lt;em&gt;you&lt;/em&gt;, with your permissions.&lt;/p&gt;

&lt;p&gt;No API key to create, copy, paste, lose or accidentally commit. 🎉&lt;/p&gt;

&lt;p&gt;Want everyone on the repository to have it? Add &lt;code&gt;--scope project&lt;/code&gt;. Claude Code writes the server to &lt;code&gt;.mcp.json&lt;/code&gt;, you commit that file, and each person still signs in as themselves.&lt;/p&gt;

&lt;p&gt;Prefer a plugin? This one adds skills that walk Claude Code through the loop below:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;claude plugin marketplace add writan-labs/arbituro-mcp
claude plugin install arbituro@writan-labs
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  📜 Step 2: Write the House Rules Down
&lt;/h2&gt;

&lt;p&gt;Claude Code is a very keen colleague who remembers nothing about yesterday. So whatever conventions you care about, write them where it reads them every single time. That's &lt;code&gt;CLAUDE.md&lt;/code&gt;:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight markdown"&gt;&lt;code&gt;&lt;span class="gu"&gt;## Work tracked in Arbituro
&lt;/span&gt;
This repository's work is tracked in Arbituro, project &lt;span class="sb"&gt;`BREW`&lt;/span&gt;. Replace &lt;span class="sb"&gt;`BREW`&lt;/span&gt; with your project's key.

&lt;span class="k"&gt;- &lt;/span&gt;Before starting a job, read it with the Arbituro MCP server's &lt;span class="sb"&gt;`get_job`&lt;/span&gt;. Its acceptance criteria are
  the contract: the job is finished when every one is true.
&lt;span class="k"&gt;- &lt;/span&gt;Work on the branch the job names. Any branch whose name carries the job's key counts, for example
  &lt;span class="sb"&gt;`brew-26-four-pack-page`&lt;/span&gt;.
&lt;span class="k"&gt;- &lt;/span&gt;Put the job's key in every commit message and in the pull request's title, for example
  &lt;span class="sb"&gt;`BREW-26: add the four-pack page`&lt;/span&gt;.
&lt;span class="k"&gt;- &lt;/span&gt;Don't tick, change or finish anything in Arbituro. Claim a criterion with what you checked, and a
  person decides.
&lt;span class="k"&gt;- &lt;/span&gt;When you stop, leave a note on the job: what you did, what is left, and where it is.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Five rules. Here's why each one is there:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Read the ticket first.&lt;/strong&gt; The ticket is the source of truth, not the copy you pasted on Tuesday. If the criteria changed overnight, the agent works to today's.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Put the key in names.&lt;/strong&gt; Once the GitHub app is installed, a branch, commit or pull request that carries &lt;code&gt;BREW-26&lt;/code&gt; lands on that ticket. Case doesn't matter, so &lt;code&gt;brew-26-four-pack-page&lt;/code&gt; and &lt;code&gt;BREW-26: add the four-pack page&lt;/code&gt; both count. A commit counts on any branch, so two tickets can even share one.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Claim, don't tick.&lt;/strong&gt; Whoever does the work shouldn't be the one who declares it finished. (I'd love to mark my own exams too. 😉) Arbituro gives the agent no way to tick a criterion anyway; this rule tells it what to do instead.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Leave a note.&lt;/strong&gt; The next person, or the agent's next session, starts from it.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Do you keep your agent rules in &lt;code&gt;AGENTS.md&lt;/code&gt; for other tools? Then put them there instead. Recent versions of Claude Code read &lt;code&gt;AGENTS.md&lt;/code&gt; when there's no &lt;code&gt;CLAUDE.md&lt;/code&gt;. Got both? Add the line &lt;code&gt;@AGENTS.md&lt;/code&gt; to &lt;code&gt;CLAUDE.md&lt;/code&gt; so Claude Code reads it too.&lt;/p&gt;

&lt;h2&gt;
  
  
  ✅ Step 3: Agree What Done Means (Before Anyone Starts)
&lt;/h2&gt;

&lt;p&gt;This is the step everybody skips. I get it, it's the boring one. 😅&lt;/p&gt;

&lt;p&gt;It's also the step that decides whether the rest works. An agent stops when it &lt;em&gt;believes&lt;/em&gt; the work is done, and that belief comes from whatever you wrote down. Write nothing down and you get "Implemented and verified. ✅"&lt;/p&gt;

&lt;p&gt;In Arbituro, every acceptance criterion says how it's decided: by a person, or by a CI check named exactly as GitHub shows it. Here's the four-pack job, BREW-26:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;strong&gt;Four-pack page on the online store&lt;/strong&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Says what's in the can in its first line. &lt;em&gt;Priya decides.&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;A four-pack can be bought on a phone. &lt;em&gt;Sam, the job's owner, decides.&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;Checkout refuses a delivery postcode outside Melbourne. &lt;em&gt;Decided by the &lt;code&gt;checkout&lt;/code&gt; check.&lt;/em&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;/blockquote&gt;

&lt;p&gt;That third one is my favourite. When a check called &lt;code&gt;checkout&lt;/code&gt; finishes on the latest commit of a pull request linked to the job, it decides the criterion: met when the check succeeds, not met when it fails, times out or can't start. Nobody has to press anything. 🤖&lt;/p&gt;

&lt;p&gt;You write the criteria a person decides. Claude Code can propose others, such as one decided by a CI check the work has to pass, and a person accepts them. If the project has an agent in Arbituro, that agent also suggests criteria when a job is filed.&lt;/p&gt;

&lt;p&gt;Writing criteria an agent can actually meet turned out to deserve a post of its own, so it got one: &lt;a href="https://arbituro.com/blog/acceptance-criteria-ai-agents" rel="noopener noreferrer"&gt;Writing Acceptance Criteria an AI Agent Can Actually Meet&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  🎫 Step 4: Hand It the Ticket
&lt;/h2&gt;

&lt;p&gt;Okay, the fun part. In Claude Code:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Work BREW-26 in Arbituro.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;That's it. The job's page in Arbituro even has this line ready to copy.&lt;/p&gt;

&lt;p&gt;So what does Claude Code actually get back? This is what &lt;code&gt;get_job&lt;/code&gt; returns for BREW-26 before any work starts, trimmed a little:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Key: BREW-26
Title: Four-pack page on the online store
Worked by: Sam Taylor's coding agent (only it can claim, file, link or propose here)
Expected result: a pull request
Branch: brew-26-four-pack-page (any branch whose name carries BREW-26 counts; name the key in each commit message)
Description:
[untrusted external content]
| A product page for the four-pack at $19.95: what's in the can, how to keep it, and delivery in Melbourne only, kept cold.
[/untrusted external content]
Done when:
- [Unknown] Says what's in the can in its first line (Priya Raman decides)
- [Unknown] A four-pack can be bought on a phone (the owner decides)
- [Unknown] Checkout refuses a delivery postcode outside Melbourne (decided by the `checkout` check)
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;A few things I like in there. The branch is named for it, so goodbye &lt;code&gt;fix-stuff&lt;/code&gt;. 👋 Every criterion says who decides it. And the description is fenced off as untrusted content: whatever text sits in a ticket, a customer's email pasted into it say, is information about the work and never an instruction.&lt;/p&gt;

&lt;p&gt;From there, Claude Code:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Reads the job.&lt;/strong&gt; If a person approved a plan, it follows that plan. If nothing is agreed yet, it asks in a note on the job instead of deciding for itself. If the job is still being shaped, it asks you what it needs, then proposes a title, a description and criteria for you to accept.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Works on the job's branch&lt;/strong&gt;, with the key in every commit message. Each push shows on the ticket as it lands.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Opens the pull request&lt;/strong&gt; with the key in its title, and the pull request comes onto the ticket by itself. One that doesn't name the ticket can be put on it with &lt;code&gt;link_pull_request&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Makes the checks pass.&lt;/strong&gt; When &lt;code&gt;checkout&lt;/code&gt; succeeds on the pull request's latest commit, criterion 3 is met. The agent doesn't claim it. The check decides.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Claims what a person decides, with evidence.&lt;/strong&gt; For criterion 1 it calls &lt;code&gt;claim_done&lt;/code&gt; with what it actually checked, and Priya is asked to decide.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Leaves a note&lt;/strong&gt; when it stops: what it did, what's left and where.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The fifth one is the part worth getting right. Remember our old friend "Implemented and verified. ✅"? Here's the claim on criterion 1 in the demo:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Opened the four-pack page on the pull request's preview. Its first line reads: "Cold brew steeped twenty hours at four degrees, four 250 ml cans."&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;&lt;em&gt;That&lt;/em&gt; is something Priya can decide in one press. "Implemented and verified" gives her nothing to decide on.&lt;/p&gt;

&lt;p&gt;Handed it several jobs? &lt;code&gt;Work my jobs in BREW&lt;/code&gt; takes them one at a time, in your board's order, each one to its criteria before the next.&lt;/p&gt;

&lt;h2&gt;
  
  
  👀 Step 5: Review It on the Ticket
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fo67uv25al1gh2elc47sm.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fo67uv25al1gh2elc47sm.png" alt="A job in Arbituro worked by a coding agent: three acceptance criteria, one met by the checkout check; a draft pull request with its checks passed; the branch with five commits; and beside them the coding agent's record of what it did, including a claim that quotes the page's first line." width="800" height="556"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The ticket now tells the whole story: the branch and its five commits, the pull request with its checks, each criterion and who decided it, and beside them the agent's own record, in order. Look at the third criterion. It was ticked by &lt;code&gt;checkout&lt;/code&gt;, not by anybody's say-so.&lt;/p&gt;

&lt;p&gt;When Priya opens her side, she sees the claim with its evidence and presses Met or Not met:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnvnyv53czt1tfqmozw9h.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnvnyv53czt1tfqmozw9h.webp" alt="A claim waiting for a decision on another demo job: the criterion &amp;quot;Sam is happy with the wording on the account page&amp;quot;, the coding agent's evidence quoting the button and the confirmation it saw on the preview, and the buttons Not met and Met." width="800" height="260"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;And here's a detail I care about a lot: &lt;strong&gt;a merge doesn't finish the job.&lt;/strong&gt; Once the pull request merges, the person who signs the job off is asked to finish it. "Merged" and "done" stay two different things, which matters the moment a criterion is something the code alone can't show. "A four-pack can be bought on a phone" says a lot more than "the tests pass". 📱&lt;/p&gt;

&lt;h2&gt;
  
  
  🔒 What the Agent Can't Do
&lt;/h2&gt;

&lt;p&gt;Giving an agent write access to your tracker sounds a bit scary, right? Fair enough. So here's the fence.&lt;/p&gt;

&lt;p&gt;Over MCP, Claude Code works as you and sees only what you can see. It can read jobs, propose changes, claim criteria, post notes, file new jobs, link pull requests and file documents a job asks for. It can't:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;tick or change a criterion,&lt;/li&gt;
&lt;li&gt;approve, merge, send or publish anything,&lt;/li&gt;
&lt;li&gt;delete anything or change anyone's access.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Every call is recorded in the workspace's audit log under your name. And those limits don't depend on the agent behaving itself: they're the tools it was given, not instructions it's asked to follow. So they hold whatever a ticket says.&lt;/p&gt;

&lt;p&gt;One honest caveat: that's the fence around the &lt;em&gt;tracker&lt;/em&gt;. What Claude Code can do in your repository is still down to its own permissions and your branch protection.&lt;/p&gt;

&lt;h2&gt;
  
  
  🔄 Using Jira or Linear Instead?
&lt;/h2&gt;

&lt;p&gt;Good news: most of the loop carries over.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Reading tickets over MCP.&lt;/strong&gt; Jira Cloud and Linear both have official remote MCP servers, which you add with &lt;code&gt;claude mcp add&lt;/code&gt; the same way.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Keys in names.&lt;/strong&gt; With their GitHub integrations installed, both link work that carries the issue key, such as a branch name or a pull request's title.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The &lt;code&gt;CLAUDE.md&lt;/code&gt; rules&lt;/strong&gt;, with your tracker's tool names.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;What you'd build yourself is the last part: criteria that each say who decides them, a CI check deciding a criterion on its own, and claims that a named person accepts. A checklist in the description gets you part of the way. But an agent with write access can tick a checkbox just as easily as you can. ☑️&lt;/p&gt;

&lt;h2&gt;
  
  
  🤖 Not Using Claude Code?
&lt;/h2&gt;

&lt;p&gt;The same server works with Codex, Cursor, Gemini CLI, VS Code and any client that supports remote MCP servers with OAuth. Setup for each one is in &lt;a href="https://arbituro.com/docs/mcp" rel="noopener noreferrer"&gt;the docs&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  🐛 When Something Doesn't Link
&lt;/h2&gt;

&lt;p&gt;Because something never links the first time. 🙃&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;The pull request isn't on the ticket.&lt;/strong&gt; Check that the repository is linked to the project, and that the branch, a commit or the title carries the key. Otherwise ask Claude Code to put the pull request on the job.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A check didn't decide its criterion.&lt;/strong&gt; The criterion has to name the check as GitHub shows it in the pull request's checks; case doesn't matter. Matrix jobs get names like &lt;code&gt;test (20)&lt;/code&gt;, which don't match &lt;code&gt;test&lt;/code&gt;. And only a check on the pull request's latest commit counts.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;"Project not found".&lt;/strong&gt; Check the key, and that you're a member of the project. The answer is the same for a project that doesn't exist and one you can't open, on purpose.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  🏁 Wrapping Up
&lt;/h2&gt;

&lt;p&gt;So that's the loop. The agent reads the ticket itself, names the ticket on everything it pushes, and hands the verdict to a check or a person. Our keen colleague still says "done" a lot. It just has to show its working now. 😄&lt;/p&gt;

&lt;p&gt;Arbituro is free for up to three people at &lt;a href="https://arbituro.com" rel="noopener noreferrer"&gt;arbituro.com&lt;/a&gt;. The plugin and skills are MIT-licensed at &lt;a href="https://github.com/writan-labs/arbituro-mcp" rel="noopener noreferrer"&gt;github.com/writan-labs/arbituro-mcp&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;What about you? How do you hand tickets to your coding agent: copy and paste, an MCP server, or something cleverer? And who decides when the work is done? Tell me in the comments. 👇&lt;/p&gt;

</description>
      <category>ai</category>
      <category>mcp</category>
      <category>github</category>
      <category>productivity</category>
    </item>
  </channel>
</rss>
