<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by AI CodeOps on Medium]]></title>
        <description><![CDATA[Stories by AI CodeOps on Medium]]></description>
        <link>https://medium.com/@aicodeops?source=rss-7ff82ca5309b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*VHxjuaA5keoLEBPg1jssGQ.jpeg</url>
            <title>Stories by AI CodeOps on Medium</title>
            <link>https://medium.com/@aicodeops?source=rss-7ff82ca5309b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 05:32:56 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@aicodeops/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com" rel="hub"/>
        <item>
            <title><![CDATA[Your repo can talk to your AI agent: prompt injection for coding teams]]></title>
            <link>https://medium.com/@aicodeops/your-repo-can-talk-to-your-ai-agent-prompt-injection-for-coding-teams-7feaa8d2a8f4?source=rss-7ff82ca5309b------2</link>
            <guid isPermaLink="false">https://medium.com/p/7feaa8d2a8f4</guid>
            <category><![CDATA[security]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[AI CodeOps]]></dc:creator>
            <pubDate>Tue, 17 Nov 2026 16:20:13 GMT</pubDate>
            <atom:updated>2026-11-17T16:20:13.604Z</atom:updated>
            <content:encoded><![CDATA[<p>A coding agent reads a lot before it writes anything: the issue it was assigned, comments on it, the README, docs, source files, test output, and sometimes web pages and package docs. It treats all of that as context.</p><p>The problem is that a language model can’t reliably tell the difference between context it should use and instructions it should follow. If text in that context says “ignore the task and do this instead,” the model may comply. This is prompt injection, and OWASP lists it as the top risk in its Top 10 for LLM applications.</p><p>For coding teams, the question isn’t whether a model can be tricked. It’s what an agent can do once it has been.</p><h3>Where injected instructions can come from</h3><ul><li><strong>Issues and comments.</strong> On public repositories, anyone can open an issue. If agents pick up issues automatically, the issue body is untrusted input.</li><li><strong>Pull request descriptions and review comments</strong> from outside contributors.</li><li><strong>Files in the repository</strong>, including docs, config files and code comments, especially in dependencies or vendored code.</li><li><strong>Tool output.</strong> Test logs, error messages and command output that include text from elsewhere.</li><li><strong>Web pages and package docs</strong> the agent fetches while researching.</li></ul><p>Instructions can be hidden: inside HTML comments, collapsed sections, long lines, or text written to look like part of the file.</p><h3>What an injected agent might try</h3><ul><li>Add a dependency, a script or a CI step that runs attacker code</li><li>Read secrets from the environment and print them into logs or a pull request</li><li>Weaken a check: skip a test, loosen a permission, remove validation</li><li>Push to a branch or merge without review</li></ul><p>All of these depend on what the agent is allowed to do. That’s where to focus.</p><h3>Limit what a fooled agent can do</h3><p>You can’t make a model immune to prompt injection. You can make injection low-impact:</p><ol><li><strong>No secrets in the agent’s environment.</strong> Run agents with the minimum credentials needed, and keep production keys out entirely.</li><li><strong>Scoped repository tokens.</strong> Access only to the repositories in the task, and no permission to merge or push to protected branches.</li><li><strong>Sandboxed execution.</strong> Tests and commands run in an isolated environment with limited network access.</li><li><strong>Treat outside input as untrusted.</strong> Don’t let agents act automatically on issues or comments from people outside the team.</li><li><strong>Review the diff, not the description.</strong> An injected change can come with a perfectly reasonable summary.</li><li><strong>Gate the merge.</strong> Nothing reaches main without tests on the exact commit, an independent review and a human approval.</li></ol><h3>How AI CodeOps handles it</h3><p>In AI CodeOps, agents work in sandboxes with tokens scoped to the repositories you choose, and policies control what each agent may do. Agents can’t approve or merge their own work. A separate AI review checks the real diff, including changes nobody asked for, and a pull request only opens when tests pass on that exact commit, the review has no open high or critical findings, and a person approves. If anything errors, the gate fails closed.</p><p>That doesn’t make prompt injection impossible. It means a fooled agent produces a pull request that gets caught, not a change that ships.</p><p><em>Free during the open beta: </em><a href="https://aicodeops.dev"><em>https://aicodeops.dev</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7feaa8d2a8f4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to review an AI-written pull request: a 10-point checklist]]></title>
            <link>https://medium.com/@aicodeops/how-to-review-an-ai-written-pull-request-a-10-point-checklist-23f57cfe6f35?source=rss-7ff82ca5309b------2</link>
            <guid isPermaLink="false">https://medium.com/p/23f57cfe6f35</guid>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[code-review]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[AI CodeOps]]></dc:creator>
            <pubDate>Fri, 13 Nov 2026 21:48:53 GMT</pubDate>
            <atom:updated>2026-11-13T21:53:25.024Z</atom:updated>
            <content:encoded><![CDATA[<p>AI-written pull requests have a particular look. The code is tidy, the names are sensible, the description is confident, and the tests are green. That polish makes them easy to approve and harder to review well.</p><p>The mistakes are still there. They just sit in different places than a rushed human change. Here’s a checklist for the ten minutes you spend on an AI-written pull request, roughly in the order worth doing them.</p><h4>Step 1: Read the task before the diff</h4><p>Start with what was asked: the issue, the prompt, the plan. Then ask whether the diff does that, all of it, and nothing else. Agents often solve a nearby problem, or the stated problem plus a “while I was here” refactor nobody requested.</p><h4>Step 2: Check the size against the request</h4><p>A one-line bug fix shouldn’t touch nine files. If the diff is much bigger than the task, ask for it to be split or trimmed before reviewing details. Large unrequested changes are where regressions hide.</p><h4>Step 3: Look at what was deleted</h4><p>Reviewers read added lines. Agents also remove things: a validation check, an error branch, a log line, a test case that was “flaky.” Scan the red lines on purpose.</p><h4>Step 4: Check the tests, not just the result</h4><p>Green tests only mean something if they test the change. Look for:</p><ul><li>Tests that assert what the code does rather than what it should do</li><li>Mocks that replace the very thing being tested</li><li>Tests that were changed or skipped to make the suite pass</li><li>No new test for a bug fix</li></ul><h4>Step 5: Follow the inputs</h4><p>Wherever the change reads input (request bodies, query strings, files, environment variables, model output), check that it’s validated, size-limited and safely encoded on the way out. Agents tend to secure the path they were asked about and leave a parallel path open.</p><h4>Step 6: Verify every new dependency</h4><p>Check that each new package exists, is the one you think it is, is maintained, and is actually needed. Models sometimes suggest package names that don’t exist or are close to popular ones, which is how typosquatting gets in.</p><h4>Step 7: Look for invented APIs and options</h4><p>A call to a method or a config option that sounds right but doesn’t exist in your version of the library may compile in a dynamic language and fail at runtime. If you don’t recognise it, check the docs.</p><h4>Step 8: Check error handling and failure paths</h4><p>Look at what happens when the network call fails, the file is missing, or the value is null. Common patterns in generated code are catch-all handlers that swallow errors and fallbacks that quietly return a default instead of failing.</p><h4>Step 9: Check secrets and config</h4><p>Look for hard-coded keys, tokens, URLs or credentials, test settings left on, and debug flags. Also check that nothing sensitive was added to logs.</p><h4>Step 10: Confirm what you’re approving</h4><p>Make sure the commit you’re approving is the one that will merge, and that the tests ran on it. If the agent pushed again after your review, look again.</p><h4>Make the checklist cheaper</h4><p>Most of this list can be pre-checked so the human review focuses on judgement:</p><ul><li>Run a separate AI review of the real diff and rank findings by severity, so the reviewer starts from a list.</li><li>Fail the change automatically when tests didn’t run on the final commit.</li><li>Flag new dependencies and deleted tests in the pull request summary.</li></ul><p>That’s how AI CodeOps works: AI agents plan, build, review and test changes on your repository, and a pull request only opens when your tests pass on the exact commit, the AI review has no open high or critical findings, and a person approves. The reviewer still makes the call; they just start with the evidence in front of them.</p><p><em>AI CodeOps is free during the open beta, and free for non-commercial open-source projects: </em><a href="https://aicodeops.dev"><em>https://aicodeops.dev</em></a></p><p><em>Originally published at </em><a href="https://aicodeops.hashnode.dev/how-to-review-an-ai-written-pull-request-a-10-point-checklist"><em>https://aicodeops.hashnode.dev</em></a><em> on November 13, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=23f57cfe6f35" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Code Review for Gitea and Forgejo: What Works Today]]></title>
            <link>https://medium.com/@aicodeops/ai-code-review-for-gitea-and-forgejo-what-works-today-159b098a2764?source=rss-7ff82ca5309b------2</link>
            <guid isPermaLink="false">https://medium.com/p/159b098a2764</guid>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[code-review]]></category>
            <category><![CDATA[github]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[self-hosted]]></category>
            <dc:creator><![CDATA[AI CodeOps]]></dc:creator>
            <pubDate>Tue, 03 Nov 2026 17:04:42 GMT</pubDate>
            <atom:updated>2026-11-03T17:06:01.500Z</atom:updated>
            <content:encoded><![CDATA[<p>Most AI code review tools are built for GitHub first, then GitLab and Bitbucket. If your team runs Gitea or Forgejo, often self-hosted for cost or control, the list gets short fast.</p><p>For example, CodeRabbit’s <a href="https://docs.coderabbit.ai/platforms/overview">platform docs</a> list GitHub, GitLab, Bitbucket and Azure DevOps (cloud and self-hosted variants), with no Gitea or Forgejo support as of November 2026.</p><p>AI CodeOps works with Gitea and Forgejo, including self-hosted servers, the same way it works with GitHub and GitLab: you connect the server with a token scoped to the repositories you choose, and AI agents plan, build, review and test changes on your real repository.</p><h3>What you get on Gitea or Forgejo</h3><ul><li><strong>AI review of the real diff</strong>, with findings sorted by severity</li><li><strong>Your own tests</strong>, run in a sandbox on the exact version of the change</li><li><strong>A release gate</strong>: no pull request until the tests pass, the review has no open high or critical findings, and a person approves</li><li><strong>A release-check API</strong> your CI can call on any PR or commit, so direct merges are covered too</li></ul><p>Free during the open beta, and free forever for non-commercial open-source projects: <a href="https://aicodeops.dev">https://aicodeops.dev</a></p><p><em>Originally published at </em><a href="https://dev.to/aicodeops_dev/ai-code-review-for-gitea-and-forgejo-what-works-today-3kb0"><em>https://dev.to</em></a><em> on November 3, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=159b098a2764" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Two Policy Rules to Give Your AI Coding Agents Before Anything Else]]></title>
            <link>https://medium.com/@aicodeops/two-policy-rules-to-give-your-ai-coding-agents-before-anything-else-cdb7c1467c46?source=rss-7ff82ca5309b------2</link>
            <guid isPermaLink="false">https://medium.com/p/cdb7c1467c46</guid>
            <category><![CDATA[security]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[ai-agent]]></category>
            <dc:creator><![CDATA[AI CodeOps]]></dc:creator>
            <pubDate>Tue, 03 Nov 2026 14:13:14 GMT</pubDate>
            <atom:updated>2026-11-03T14:19:11.327Z</atom:updated>
            <content:encoded><![CDATA[<p>Most teams adopting AI coding agents start with trust settings: what can the agent see, what can it change. Before you tune anything else, we’d start with two rules.</p><h3>Rule 1: No production deploys on an agent’s word</h3><pre>{ &quot;id&quot;: &quot;no-production-deploys&quot;, &quot;action&quot;: &quot;release.deploy&quot;, &quot;resource&quot;: &quot;production&quot;, &quot;decision&quot;: &quot;deny&quot; }</pre><p>An agent can propose a change, run tests and open a pull request. Shipping to production should still be a human decision, made with the evidence in front of them. A deny rule makes that explicit instead of hoping nobody wires it up.</p><h3>Rule 2: Any change to your CI files waits for a person</h3><pre>{ &quot;id&quot;: &quot;review-ci-changes&quot;, &quot;action&quot;: &quot;diff.apply&quot;, &quot;resource&quot;: &quot;.github/**&quot;, &quot;decision&quot;: &quot;ask&quot; }</pre><p>Your CI configuration is what decides whether everything else is checked. A change to .github/workflows can quietly skip a test job or loosen a branch protection step. That makes CI files the highest-leverage target in the repo, for a buggy agent or a prompt injection alike. ask holds the change on an Approvals page until someone reviews it.</p><h3>How the rules behave</h3><ul><li><strong>Strictest wins.</strong> When several rules match, deny beats ask, and ask beats allow, whatever their order.</li><li><strong>Ask fails closed.</strong> Where an action can’t wait for a person, an ask is enforced as a deny.</li><li><strong>One reject stops it.</strong> On the Approvals page, a single reject ends the request.</li><li><strong>Versioned.</strong> Each save creates a new policy version, so you can see what changed and when.</li></ul><p>If you’re unsure about a rule’s impact, run it in log-only mode first and watch which actions it would have blocked.</p><p>Video walkthroughs are on our YouTube channel: <a href="https://www.youtube.com/@AICodeOps">https://www.youtube.com/@AICodeOps</a></p><p>AI CodeOps is free during the open beta and free forever for non-commercial open source: <a href="https://aicodeops.dev">https://aicodeops.dev</a></p><p><em>Originally published at </em><a href="https://dev.to/aicodeops_dev/two-policy-rules-to-give-your-ai-coding-agents-before-anything-else-eh"><em>https://dev.to</em></a><em> on November 3, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cdb7c1467c46" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Our AI’s Security Audit Flagged a Bug We Never Asked It to Fix]]></title>
            <link>https://medium.com/@aicodeops/our-ais-security-audit-flagged-a-bug-we-never-asked-it-to-fix-f6cbcd483593?source=rss-7ff82ca5309b------2</link>
            <guid isPermaLink="false">https://medium.com/p/f6cbcd483593</guid>
            <category><![CDATA[devsecops]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[code-review]]></category>
            <dc:creator><![CDATA[AI CodeOps]]></dc:creator>
            <pubDate>Tue, 03 Nov 2026 14:12:49 GMT</pubDate>
            <atom:updated>2026-11-03T14:16:18.908Z</atom:updated>
            <content:encoded><![CDATA[<p>We gave a set of AI agents a small, realistic task on a demo shopping cart: discount codes now come from user input, so validate them. Trim whitespace, uppercase, letters and digits only, 20 characters max, and reject anything else before it reaches the totals. Add tests.</p><p>We ran it through a Security-Sensitive Change workflow: five agents in a row. One scopes the change, one analyses the risk, one implements, one audits the diff, and one validates the result. The whole run took about 35 seconds and cost under 20 cents.</p><p>The implementation did what we asked. The diff added a validateDiscountCode helper that trims, uppercases, checks the pattern and length, and throws on anything else. Tests covered valid codes, illegal characters, long codes and unknown codes.</p><p>Then the audit step came back with four findings. Three were minor. One was marked HIGH:</p><blockquote><em>Unvalidated </em><em>discountPercent parameter allows arbitrary discount values.</em></blockquote><p>The cart’s older total() function still accepted a raw discountPercent option. We had asked the agents to lock down the new input path, and they did. But the old path was still wide open: a caller could pass 100, or 250, and get a free cart or a negative total. Nobody had asked about it. The auditor found it anyway, because it was reading the code around the change, not just the change.</p><p><strong>The gate held.</strong> Because a HIGH finding was open, the change was not release-ready. It could not become a pull request until the finding was fixed or a person signed off on it. One click on Fix sends the finding back to an agent as a new task.</p><h3>Three takeaways</h3><ol><li><strong>Audit the code around the change, not just the diff lines.</strong> The most dangerous bugs were next to what we changed, not in it.</li><li><strong>A finding needs teeth.</strong> If a reviewer can flag a HIGH issue and the change still merges, the review is advice. Tie open high or critical findings to a merge block.</li><li><strong>Scope creep is good when it’s a finding, not a commit.</strong> The auditor didn’t silently “fix” the extra issue inside our change. It reported it, so a person decides.</li></ol><p>We recorded runs like this as short tutorials on our YouTube channel: <a href="https://www.youtube.com/@AICodeOps">https://www.youtube.com/@AICodeOps</a></p><p>AI CodeOps is free during the open beta and free forever for non-commercial open source: <a href="https://aicodeops.dev">https://aicodeops.dev</a></p><p><em>Originally published at </em><a href="https://dev.to/aicodeops_dev/our-ais-security-audit-flagged-a-bug-we-never-asked-it-to-fix-4lap"><em>https://dev.to</em></a><em> on November 3, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f6cbcd483593" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Coding Agents Shouldn’t Write Surprise Invoices]]></title>
            <link>https://medium.com/@aicodeops/ai-coding-agents-shouldnt-write-surprise-invoices-4a73a0350f70?source=rss-7ff82ca5309b------2</link>
            <guid isPermaLink="false">https://medium.com/p/4a73a0350f70</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[engineering-management]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[AI CodeOps]]></dc:creator>
            <pubDate>Mon, 02 Nov 2026 01:00:12 GMT</pubDate>
            <atom:updated>2026-11-02T01:00:12.270Z</atom:updated>
            <content:encoded><![CDATA[<h4>How to keep AI coding costs predictable: cost per session, budgets that actually stop spend, and caps per workflow</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*273YNO_tpK7gPWP6kcgyrQ.png" /></figure><p>AI coding agents are billed by the token, and agents use a lot of tokens. A planning step, a build step, a review, a test-fixing loop: each one calls a model, sometimes many times. One long session can cost more than a week of short ones, and most teams only find out when the invoice arrives.</p><p>Predictable AI spend comes down to three things: seeing the cost, limiting it, and knowing what happens at the limit.</p><h3>1. See the cost of every session</h3><p>You can’t manage what you only see monthly. In AI CodeOps, every model call is recorded and rolled up:</p><ul><li><strong>Per session:</strong> each session shows its total cost, plus a table of every agent step with its runtime, cost and tokens.</li><li><strong>Per change:</strong> each proposed change shows what the session that produced it spent.</li><li><strong>Per day and model:</strong> the Usage page charts daily spend by model, filterable by model and API key.</li></ul><p>Tokens from failed model calls count too, because the provider charges for them. A cost view that skips failures will always look cheaper than the bill.</p><h3>2. Set a budget that actually stops spend</h3><p>A budget that only sends a warning is a report, not a limit. AI CodeOps budgets have two levels:</p><ul><li><strong>Soft limit:</strong> a marker. Past it, starting a session shows a warning with the period’s spend, and the crossing is recorded.</li><li><strong>Hard limit:</strong> new sessions are refused, and running sessions stop once spend reaches it.</li></ul><p>Budgets can run daily, weekly, monthly or on a rolling 30 days. The budget is checked before every model call, so a runaway session stops close to the limit instead of at the end of the month.</p><h3>3. Cap the expensive workflows separately</h3><p>Not all work costs the same. A full feature workflow with planning, building, multiple reviews and tests costs much more than a quick fix. Per-workflow caps let you keep the expensive one within its own limit while quick ones run freely under the workspace budget.</p><h3>Know what happens at the edges</h3><p>Good controls are honest about their edges:</p><ul><li>If the budget check fails when a session starts, the session is blocked rather than run unmetered.</li><li>A single model call already in flight can take spend slightly past the limit.</li><li>Budgets apply whether a session runs on the platform’s models, on credits, or on your own provider key.</li></ul><h3>Bring your own key, keep the same controls</h3><p>Many teams want to use their own model provider account for billing or data reasons. That shouldn’t mean giving up spend controls. Sessions on your own key show the same cost and respect the same budgets.</p><h3>The takeaway</h3><p>Ask your AI coding tool three questions. What did that session cost? What stops a session that’s running away? Can I cap an expensive workflow separately? If the answers are “check the invoice”, “nothing” and “no”, the surprise is already scheduled.</p><p>AI CodeOps is free during the open beta: <a href="https://aicodeops.dev">aicodeops.dev</a></p><p><em>Drafted with AI assistance and edited by the AI CodeOps team.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4a73a0350f70" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your Tests Passed. On Which Version?]]></title>
            <link>https://medium.com/@aicodeops/your-tests-passed-on-which-version-9da9a9bffd4d?source=rss-7ff82ca5309b------2</link>
            <guid isPermaLink="false">https://medium.com/p/9da9a9bffd4d</guid>
            <category><![CDATA[software-testing]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[code-review]]></category>
            <category><![CDATA[software-engineering]]></category>
            <dc:creator><![CDATA[AI CodeOps]]></dc:creator>
            <pubDate>Mon, 02 Nov 2026 00:58:32 GMT</pubDate>
            <atom:updated>2026-11-02T00:58:32.029Z</atom:updated>
            <content:encoded><![CDATA[<h4>Why a test result should only count for the exact change it ran on, especially when AI writes the code</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FiGbd1IrMhbvwG1A9W_fMg.png" /></figure><p>Here’s a small thing that causes real incidents.</p><p>An AI agent writes a change. The test suite runs and passes. Someone notices a naming issue and tweaks the change by hand, or the agent is asked for one more small fix. The pull request still shows a green check, so it merges.</p><p>But the green check belongs to the version before the edit. Nobody tested what actually shipped.</p><h3>Why this gets worse with AI</h3><p>With people, edits after review tend to be few and visible. With AI agents, a change can go through several quick rounds: generate, review, fix, adjust. Each round is a new version. If your process treats “tests passed at some point” as “tests passed”, the faster the iteration, the wider the gap between what was tested and what merges.</p><p>It’s a quiet failure. Nothing looks wrong. The check is green, the approval is there, and the diff looks reasonable.</p><h3>The rule: a result belongs to a version</h3><p>The fix is a simple rule. A test result counts only for the version it ran on.</p><p>In AI CodeOps, that rule is built into the release gate:</p><ul><li>After an agent writes a change, the session runs the project’s own test command in a sandbox and records the result against that version.</li><li>If anyone edits the change after its tests ran, the old result <strong>neither passes nor blocks it</strong>. The gate marks the new version as untested.</li><li>The approval opens again too, so a person decides before the edited version can ship.</li><li>Re-running the session tests the new version and the gate can clear it.</li></ul><p>Notice the second point. The old result doesn’t count as a failure either. It simply stops counting, because it’s evidence about a different change.</p><h3>The same rule in CI</h3><p>The gate your CI can call follows it too. It looks at every session that touched the pull request or commit. You can also ask it to deny while a session is still running (blockOnInFlight), so a merge can&#39;t sneak in between a new version and its test run.</p><h3>What this means for your team</h3><p>You don’t need a special tool to apply the principle. Ask three questions about any green check before merging:</p><ol><li>Which commit did this run on?</li><li>Has the change been edited since?</li><li>Does the approval cover the version I’m merging, or an earlier one?</li></ol><p>If you can’t answer all three from your tooling, your process has the gap. With AI writing more of the code, it’s worth closing.</p><p>AI CodeOps is free during the open beta: <a href="https://aicodeops.dev">aicodeops.dev</a></p><p><em>Drafted with AI assistance and edited by the AI CodeOps team.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9da9a9bffd4d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Free Forever for Open Source: Why We’re Giving AI CodeOps to Maintainers]]></title>
            <link>https://medium.com/@aicodeops/free-forever-for-open-source-why-were-giving-ai-codeops-to-maintainers-50834221c2d5?source=rss-7ff82ca5309b------2</link>
            <guid isPermaLink="false">https://medium.com/p/50834221c2d5</guid>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[github]]></category>
            <dc:creator><![CDATA[AI CodeOps]]></dc:creator>
            <pubDate>Fri, 30 Oct 2026 02:02:09 GMT</pubDate>
            <atom:updated>2026-10-30T02:03:03.214Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Maintainers carry a lot of the software world. If AI agents are going to touch their projects, they should do it behind the same quality gates as everyone else, at no cost.</em></p><p>Open-source maintainers are in an odd spot right now. AI tools make it easy for anyone to send a pull request, so the volume of contributions goes up. The time a maintainer has to review them does not.</p><p>The usual answers are to review faster or to say no more often. Neither scales. What does scale is making every change prove itself before a person spends time on it.</p><p>That’s the idea behind AI CodeOps, and it’s why we’re making it <strong>free forever for non-commercial open-source projects</strong>.</p><h3>What maintainers get</h3><p>The open source plan isn’t a trial. Approved projects get, at no cost:</p><ul><li><strong>Higher limits than the free plan:</strong> more workspaces, seats and AI models</li><li><strong>300 sessions a month</strong> on the platform’s models. Past that, you can keep going on your own model provider key, or wait for the monthly reset</li><li><strong>The release-gate badge</strong> for your README, turned on for you</li><li><strong>A usage meter</strong> in the console, so you always know where you stand</li></ul><p>It’s the same product, with the same gates. Nothing is held back for paying customers when it comes to what a change must pass before it merges.</p><h3>What the gate does for a project</h3><p>Every AI change in AI CodeOps waits at a release gate. It becomes a pull request only when:</p><ol><li><strong>Your own test suite passed</strong> in a sandbox, on that exact version of the change</li><li><strong>The AI review of the real diff</strong> has no open high or critical findings</li><li><strong>A person approved it.</strong> The person who started the session can’t cast the approving vote while someone else in the workspace can</li></ol><p>If any check errors, the gate fails closed. For a maintainer, that means the pull request in your queue has already been tested and reviewed, and someone has looked at it.</p><p>Each change also carries a signed evidence pack: the task, the models used, the findings, the test runs and the approvals. When someone asks how a change got in, there’s a record.</p><h3>Your Git host, your server</h3><p>A lot of open source doesn’t live on GitHub. AI CodeOps works with GitHub, GitLab, Bitbucket, Gitea and Forgejo, including self-hosted GitLab, Gitea and Forgejo servers connected with a personal access token. Branches and pull requests stay on your server.</p><p>One honest note: if your project is on <strong>Codeberg</strong>, read its Terms of Use first. Codeberg doesn’t allow projects that mostly consist of code written by generative AI tools. The connection works, but the rules are Codeberg’s to set.</p><h3>Who qualifies</h3><ul><li>A <strong>public repository</strong> under an <strong>OSI-approved license</strong></li><li>A <strong>non-commercial</strong> project: not sold, not the core of a paid product, and not run by a company to make money from it. Community projects, libraries and tools qualify, even with sponsors or donations</li><li><strong>Approved by application.</strong> We read every one, usually within a few working days, and email you either way</li></ul><h3>How to apply</h3><ol><li>Sign up at <a href="https://aicodeops.dev">aicodeops.dev</a> with the email you’ll use for the project. Approval moves that account’s own organization onto the open source plan</li><li>Fill in the short form at <a href="https://aicodeops.dev/pricing/open-source"><strong>aicodeops.dev/pricing/open-source</strong></a>: email, repository URL, maintainer’s GitHub username, license, and a few words about the project</li></ol><p>And as with every plan: AI CodeOps doesn’t use your code, prompts or outputs to train AI models.</p><p>If you maintain something people depend on, we’d like to help you ship changes you can trust. Apply here: <a href="https://aicodeops.dev/pricing/open-source">aicodeops.dev/pricing/open-source</a>. Program updates are posted at <a href="https://github.com/aicodeops-dev/announcements/discussions">github.com/aicodeops-dev/announcements</a>.</p><p><em>Drafted with AI assistance and edited by the AI CodeOps team.</em></p><p><em>Originally published at </em><a href="https://dev.to/aicodeops_dev/free-forever-for-open-source-why-were-giving-ai-codeops-to-maintainers-3d3f"><em>https://dev.to</em></a><em> on October 30, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=50834221c2d5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Coding Agents Beyond GitHub: GitLab, Bitbucket, Gitea and Forgejo]]></title>
            <link>https://medium.com/@aicodeops/ai-coding-agents-beyond-github-gitlab-bitbucket-gitea-and-forgejo-de5f40991335?source=rss-7ff82ca5309b------2</link>
            <guid isPermaLink="false">https://medium.com/p/de5f40991335</guid>
            <category><![CDATA[gitlab]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[self-hosted]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[AI CodeOps]]></dc:creator>
            <pubDate>Fri, 30 Oct 2026 02:02:05 GMT</pubDate>
            <atom:updated>2026-10-30T02:04:06.023Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Most AI coding agents assume your code lives on GitHub. A lot of it doesn’t. Here’s what it takes to run gated AI changes on the Git host you already use.</em></p><p>If you read the docs of most AI coding agents closely, one line keeps showing up: works with repositories hosted on GitHub. That’s a reasonable place to start. GitHub is huge.</p><p>But plenty of teams aren’t there. Enterprises run self-managed GitLab. Agencies live in Bitbucket. Privacy-minded teams and a growing slice of open source run Gitea or Forgejo on their own servers. For them, “just use the agent” means moving their code first, which usually means not using the agent.</p><p>We built AI CodeOps to work where the code already is.</p><h3>One workflow, five Git hosts</h3><p>AI CodeOps connects to <strong>GitHub, GitLab, Bitbucket, Gitea and Forgejo</strong>. The flow is the same on each:</p><ol><li><strong>Connect once.</strong> Use OAuth for the hosted services, or a personal access token for your own server. For a self-hosted GitLab, Gitea or Forgejo, you enter the server’s public https address and paste a token</li><li><strong>Describe a change in the browser.</strong> Agents plan it, write it, review it and run the project’s own tests in a sandbox</li><li><strong>Get a pull request on your host.</strong> When the change passes the release gate, AI CodeOps pushes a branch named codeops/run-&lt;id&gt; and opens a pull request (a merge request on GitLab) into the default branch</li></ol><p>A team that uses more than one host can run all of them from one workspace.</p><h3>What it does, and what it doesn’t</h3><p>For teams that own their tooling, the second list matters as much as the first. On Gitea and Forgejo, for example, AI CodeOps:</p><ul><li><strong>Does</strong> list the repositories, branches, files and pull requests the connected account can see</li><li><strong>Does</strong> push a branch and open a pull request, with commits authored as “AI CodeOps”</li><li><strong>Doesn’t</strong> need admin access. An account that can read the repository, push a branch and open pull requests is enough</li><li><strong>Doesn’t</strong> install webhooks, post commit statuses or comment on your pull requests</li><li><strong>Never</strong> pushes to your default branch directly</li></ul><p>Tokens are encrypted at rest and never written into a repository’s Git configuration. To remove access, disconnect in AI CodeOps or revoke the token on your server.</p><h3>The gate goes with you</h3><p>The point isn’t only that the agent can reach your host. It’s that the same rules apply there. A change opens a pull request only when your tests passed on that exact version, the AI review of the real diff has no open high or critical findings, and a person approved it. If a check errors, the gate fails closed.</p><p>A pull request opened by AI CodeOps can be merged from the console only after the release gate passes again. You can also merge it on your host as usual, and if you want those merges gated too, your CI can ask the gate for a verdict through an API and fail the job on a “no”.</p><h3>A note for Codeberg users</h3><p>Codeberg runs Forgejo, so the connection works with https://codeberg.org and a personal access token. But read Codeberg&#39;s Terms of Use first: they don&#39;t allow projects that mostly consist of code written by generative AI tools. That&#39;s their call, and worth respecting.</p><h3>Try it on your host</h3><p>AI CodeOps runs in the browser, with nothing to install. It’s free during the open beta, and <strong>free forever for non-commercial open-source projects</strong>.</p><p><em>Drafted with AI assistance and edited by the AI CodeOps team.</em></p><p><em>Originally published at </em><a href="https://dev.to/aicodeops_dev/ai-coding-agents-beyond-github-gitlab-bitbucket-gitea-and-forgejo-52f6"><em>https://dev.to</em></a><em> on October 30, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=de5f40991335" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your Merge Button Is the Real Quality Gate]]></title>
            <link>https://medium.com/@aicodeops/your-merge-button-is-the-real-quality-gate-5e0e937858b2?source=rss-7ff82ca5309b------2</link>
            <guid isPermaLink="false">https://medium.com/p/5e0e937858b2</guid>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[continuous-integration]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[AI CodeOps]]></dc:creator>
            <pubDate>Thu, 29 Oct 2026 21:49:42 GMT</pubDate>
            <atom:updated>2026-10-30T01:45:29.563Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A release gate only works if the merge button has to ask it. Here’s how to wire one into CI so it fails closed.</em></p><p>In the last post, <a href="https://dev.to/aicodeops_dev/nothing-should-merge-on-an-ais-word-23b8">Nothing Should Merge on an AI’s Word</a>, I argued that AI-written code needs a gate between “the agent finished” and “this is a pull request,” and that the gate should check evidence: a review that ran, tests that passed on this exact version, and a person who said yes.</p><p>There’s a hole in that picture. A gate inside one tool only covers changes made through that tool. Pull requests still get opened by hand, branches still get pushed, and the merge button on GitHub or GitLab doesn’t know your AI platform exists. If the gate can be skipped by clicking somewhere else, it’s not a gate.</p><p>So the real question is: <strong>how do you make the merge itself ask for a verdict?</strong></p><h3>Put the check where the merge happens</h3><p>The answer is boring, which is a good sign: a CI job that asks the gate for a verdict on this pull request, and fails when the answer is no. Then you make that job a <strong>required status check</strong> in branch protection.</p><p>That second step is the one teams forget. A CI job that isn’t required is a dashboard. It only becomes a gate when the merge button is greyed out until it passes.</p><p>In AI CodeOps, the question goes to one endpoint, POST /api/release/check, with a workspace API key. You send a pull request URL (or a commit SHA, at least 7 characters) and get back a verdict that covers every AI session that touched that change:</p><ul><li>summary and reasons: plain-English explanations</li><li>blockingFindings, testFailures, coverageShortfall, licenseViolations: the detail behind a deny</li><li>runsConsidered: which sessions the verdict was based on</li></ul><h3>The trap: “200 OK” doesn’t mean “allowed”</h3><p>Here’s the detail that makes or breaks a CI gate. The endpoint returns <strong>HTTP 200 for both outcomes</strong>. A denied change is a successful API call with &quot;allowed&quot;: false in the body.</p><p>That means the most common way to write this check is wrong:</p><p>curl -f only fails on HTTP errors. A deny comes back as 200, so the step goes green and the change merges. That&#39;s a gate that fails open, which is the one thing a gate must never do.</p><p>The fix is to read the body and treat anything other than an explicit true as a no:</p><p>jq -e exits non-zero when the result is false or null, so a deny fails the job, and so does a missing or malformed field.</p><h3>A fail-closed GitHub Actions job</h3><p>Putting it together:</p><pre>name : ai-release-gate on : pull_request : jobs : release-gate : runs-on : ubuntu-latest steps : - name : Ask the release gate for a verdict env : WORKSPACE_API_KEY : ${{ secrets.AICODEOPS_API_KEY }} PR_URL : ${{ github.event.pull_request.html_url }} run : | set -euo pipefail body=$(jq -n --arg pr &quot;$PR_URL&quot; &#39;{ pullRequestURL: $pr, blockingSeverity: &quot;medium&quot;, minCoveragePct: 80, blockOnInFlight: true }&#39;) resp=$(curl -sSf --retry 3 --retry-all-errors \ -X POST &quot;https://console.aicodeops.dev/api/release/check&quot; \ -H &quot;Authorization: Bearer $WORKSPACE_API_KEY&quot; \ -H &quot;Content-Type: application/json&quot; \ -d &quot;$body&quot;) echo &quot;$resp&quot; | jq -r &#39;.summary&#39; echo &quot;$resp&quot; | jq &#39;.reasons&#39; # Anything other than an explicit &quot;true&quot; fails the job. echo &quot;$resp&quot; | jq -e &#39;.allowed == true&#39; &gt; /dev/null</pre><p>Every way this can go wrong ends in a red check:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*Y38q08vUUElQx5N3.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*CKeCstzTe_FW76Dh.png" /></figure><p>Gate couldn’t evaluate</p><p>allowed: false, summary &quot;Evaluation failed.&quot;</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*Jhx1RVLVRhMSRpu9.png" /></figure><p>No AI session touched this change</p><p>Denied: there’s nothing to judge</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*hVGri0N8IRiDFr5z.png" /></figure><p>Bad or missing API key (401/403)</p><p>curl -f fails the job</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*vAF2k_Abiz18sg-q.png" /></figure><p>Invalid request (400)</p><p>curl -f fails the job</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*QWA3zxwmxbkw3We6.png" /></figure><p>Rate limited (429, 60 checks/min per workspace)</p><p>curl retries, then fails if it still can’t get through</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*AirgdlTqjqsxRFBb.png" /></figure><p>Network down, API unreachable</p><p>curl retries, then fails</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*9WYQPfZLQ8p1BLAU.png" /></figure><p>Response isn’t valid JSON</p><p>jq errors, set -e fails the job</p><p>Nothing on that list ends in a green check. That’s the property you want.</p><p>Two practical notes. The key only needs the workspace:read permission, so don&#39;t hand CI a more powerful one. And use the same host you open the console on.</p><h3>Tune the bar per pipeline</h3><p>The defaults block on open high and critical findings and on failing tests. CI is where you can be stricter, because different branches deserve different bars:</p><ul><li><strong>blockingSeverity: &quot;medium&quot;</strong> blocks on medium findings too. Reasonable for anything that touches auth, payments, or infrastructure.</li><li><strong>minCoveragePct</strong> sets a coverage floor for the latest run that reported coverage.</li><li><strong>requiredWorkflows</strong> lists workflows that must have completed on this change, such as a security audit.</li><li><strong>disallowedLicenses</strong> checks the project&#39;s SBOM for licenses you can&#39;t ship.</li><li><strong>requireDeployedTo: &quot;staging&quot;</strong> denies until the change has been deployed to staging. Useful on the job that promotes to production, not on every pull request.</li><li><strong>blockOnInFlight: true</strong> denies while a session for the change is still running, so nobody merges halfway through a fix.</li></ul><p>The gate looks back 14 days by default ( lookbackMs changes that). An old session isn&#39;t silently reused forever.</p><h3>Then keep the receipt</h3><p>The gate decides whether a change ships. The evidence pack is how you prove later why it did.</p><p>Every change gets a signed pack: the task and the models used, each run with its tokens and cost, review findings and how they were triaged, test results and coverage, policy decisions, who was eligible to approve and how each person voted, and the pull request and deployment it ended up in. The code itself isn’t included. Instead the pack carries a SHA-256 hash of the diff, so you can match it against the pull request.</p><p>The part I like is that you don’t have to trust the platform to check it. Packs are signed with Ed25519 over canonical JSON, and the public keys are published at /api/evidence?action=keys. You can verify a pack in the browser at <a href="https://aicodeops.dev/evidence/verify">aicodeops.dev/evidence/verify</a> without an account, or offline with Node&#39;s built-in crypto module. Keep a copy of the keys file next to your audit records, and verification doesn&#39;t depend on the vendor still being reachable in three years.</p><p>Retired keys stop signing new packs but are never deleted, so old packs keep verifying.</p><h3>The checklist</h3><p>If you’re adding AI agents to a repository this quarter, here’s the short version:</p><ol><li><strong>Gate the merge, not just the agent.</strong> Run the verdict check in CI and make it a required status check.</li><li><strong>Read the verdict, not the status code.</strong> allowed == true or the job fails.</li><li><strong>Make every error a no.</strong> Bad key, timeout, rate limit, unreadable response: all red.</li><li><strong>Raise the bar where it matters.</strong> Stricter severity and coverage on sensitive paths, a staging requirement before production.</li><li><strong>Keep evidence you can verify yourself.</strong> A signed record beats a chat log.</li></ol><p>AI agents can write a lot of code. The merge button is where you decide how much of it you trust.</p><p>AI CodeOps is free during the open beta and <strong>free forever for non-commercial open-source projects</strong>. The full field list is in the <a href="https://aicodeops.dev/docs/api">API docs</a>, and the gate itself is described at <a href="https://aicodeops.dev/product/quality-gates">aicodeops.dev/product/quality-gates</a>.</p><p><em>Drafted with AI assistance and edited by the AI CodeOps team.</em></p><p><em>Originally published at </em><a href="https://dev.to/aicodeops_dev/your-merge-button-is-the-real-quality-gate-11l3"><em>https://dev.to</em></a><em> on October 29, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5e0e937858b2" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>