From Vibe Slop to AgentOps, Postman AI, Bug Report to Release Sign Off and More!

What if every bug report came with session replay, console errors, and network requests attached automatically? Postman just launched an AI engineer that runs QA on every pull request. Helpful or hype? Plus the free Playwright Step Decorator that's blowing up on LinkedIn. What is it? Find out in this episode of the TestGuild News Show for the week of June 8th. So grab your favorite cup of coffee or tea and let's do this. All right. First up is a new tool I just heard about. This week's show is actually brought to you by Bugsee. Which is a new platform in beta tackling everything that happens after a bug gets reported, and they're looking for your feedback. The folks at Bugsee told me that they built a bug reporting and QA workflow platform built for teams shipping web applications. Their pitch is that most teams don't struggle to find bugs, they struggle with everything after the report, like missing technical context, can you reproduce these loops, and release sign-off guesswork. And here's what it does. When a bug is reported directly from the website under test, Bugsee automatically captures the technical details developers need: annotated screenshots, screen recordings, session replay, console errors, network and API requests, plus browser, device, and environment information. Every issue is automatically tied to the correct environment release so teams can see what's blocking a deployment, what's been tested, and what's ready to ship. There's also a QA sign-off where QA leads review and approve release readiness Plus integration with all the popular tools like Jira, Slack, Azure, DevOps, and Trello. Installation runs through a snippet, an SDK, or a Chrome extension. And the beta includes a 30-day free trial, up to 90 days free for active teams. Also make sure to check out the live demo option, no email required, using that link down below and tell them Joe sent you. Next up is from one of our Automation Guild community members, and it's from Sebastian, who just released version 2.0 of his Cypress Schema Validator. Is open source plugin for validating API responses against schemas. And the plugin supports a lot of different formats like plain JSON schemas, OpenAPI 3, Swagger 2 documents through the AGV validator, plus Zod schemas through the Zod validator, all chained directly onto cy.request. Also, version 2.0 adds support for the new environment variable handling introduced in Cypress 15.10, 'cause the old cypress.environment approach is deprecated and no longer supported in Cypress 16. Staying hands-on, here's a community plugin blowing up on LinkedIn this week, over 160 reactions, because it's fixing something every Playwright page object model suffers from. What is it? Let's check it out. And this is from Ronald, who's a principal consultant in AI enabling test automation, has published Playwright Step Decorator that wraps your page object methods in Playwright's test.step automatically. The goal, in his words, test reports that describe business actions instead of implementation details without the repetitive test.step wrappers that pile up as your framework grows. So how does it work? Well, you decorate a method with a description like login as username. The placeholders pull in the actual argument values at runtime. It supports named parameters, nested object properties, and index-based references. And the detail testers are really loving in the comments is it captures the real call site. So your HTML report, TraceViewer, and CI logs point to the line in your test where the method was called, not the descriptor's internals. There's also a get step info function for attaching files and logs to the current step.. So if your current Playwright reports read like a stack trace instead of a test narrative, you need to check this out using those links down below. All right, so next up is a question every team is probably wrestling with right now. How much do you actually let an AI agent do on its own? Well, here's one example by Otter, who is a senior software engineer, and he just published a full source of a skill. It's worth studying less for what it does, but as a good example of what you can do with this type of technology. It's an agent skill that handles ticket work in your tracker, say Jira, GitHub Issues, whatever you use. But the design principle is the headline: the agent drafts, the human decides, and nothing gets written without your explicit approval. And here's how it works. You give it a screenshot or a log, it reads it, pulls timestamps, error fingerprints, and session IDs, then runs a duplicate check before it drafts anything. Then it stops, it shows you the full draft, loops through your edits, and only commits when you say submit. And the house rules are the really interesting part here. No suggested fix sections because that quietly steers wherever it picks it up towards an answer before they even looked, and a hard rule that the agent can't make a claim about a model or a library it didn't actually verify in the session. And that confirmation gate is the pattern to copy here, whether it's filing tickets, running tests, or anything else you hesitate to hand off blind to an AI agent. So next up is what I think is one of the biggest launches I've seen in the week. Postman wants an AI agent running QA on every pull request and their reasoning, something they call context debt. Matters to your job. Let's check it out. So last week, Postman has launched their AI Engineer. The post argues that as AI speeds up software production, teams accumulate context debt— the growing body of API services and contracts that every future change has to understand. And he calls this vibe slop the visible symptom, the context debt the underlying disease. For testers, the standout workflow is API testing and QA in every pull request. Plugged into the PR review process, the agent checks changes for contract regressions, breaking changes, security issues, and design inconsistencies before they merge. It can also explore undocumented APIs, produce documentation and working collectors, and run root cause analysis with reproduction steps. And it's powered by the context graph, a continuously updated map of the APIs across your Postman organization. And it runs in a sandbox cloud environment that never touches production directly. Also, write operations require explicit human approval, and Postman compares the risk model to a junior engineer whose work goes through code reviews. If you're doing anything with Postman, probably something definitely check out once again using the link down below. Next up is our follow the money segment. So Harness has acquired Codecov from Sentry, which they announced early last week, and this is aimed at embedding code coverage intelligence directly into the delivery pipeline. The framing is this: AI has teams producing more code than ever, increasingly with agents generating changes alongside humans. But the system responsible for testing and governing software hasn't kept pace. Here's the tester's angle. I think you know about Codecov shows how much of a codebase has been tested and whether it meets the threshold required to ship safely. So inside Harness, this code coverage feeds the software delivery knowledge graph, joining deployment history, security analysis, incident telemetry, and change history so coverage signals can inform real-time release decisions before code reaches production. So if you're being asked to test AI agents, don't know where to start, this next article might help. So AWS has published a guide for AgentOps, its term for the operational discipline of running AI agents in production, built around Amazon Bedrock Agent Core. The post organizes the workflow into 4 pillars: governance and security, build and operations, evaluation, and observability. So for testers, evaluation is the headline, I think. AWS says agents should be evaluated at 4 levels: individual tool, the conversation turn, the session outcomes, and the full system. And pre-production automated tests run across several dimensions, including regression, security, agentic AI, and responsible AI. And the QA engineers, along with domain experts, are named as the people who approve promotion to production. And these evaluations are critical because builds don't ship until it passes. And in production, what I think is cool, continuous online evaluation samples live traffic to catch quality drops before users do. All right, so here are two quick hits. This one's from Bengt, and he talks about how he created an essay called Machine vs. Human: Competence Development. His point: we feed AI systems documentation standards and best practices while testers get 60 mandatory compliance courses and almost no structured education on how to actually test. He argues two decades of ad hoc approaches disguised as agile transformation did real damage and warns that professionals who never learned the fundamentals may find themselves displaced sooner than they expected. So he's calling for an era of competence. Take the same care we put into teaching machines and apply it to people. I'm starting to see this as well as I go across the country at TestGuild IRLs. We're starting to age out and the new crop of testers I don't think have been tested on the fundamentals. Definitely a worthwhile read as well. And before we go, one more. This was flagged this week on LinkedIn by Matthew Skelton, and it's all about how security research is a safe deep— a worm variant that skips the npm registry entirely and pushes malicious code straight into GitHub source repos. Including an official Microsoft Azure repo. And this article goes over how it works. The commit plants a payload wired to run automatically through Cloud Code, Gemini CLI, Cursor of VS Code, and the npm test script. So as we create more and more AI code, I think security should be top of mind for all testers. And here's just one more example. All right, for links of everything of value we covered in this news episode, head on over to those links down below. So that's it for this episode of the TestGuild News Show. I'm JoeBot. My mission is to help you succeed in creating end-to-end full-stack pipeline automation awesomeness. And as always, test everything and keep the good. Cheers.