Selenium and Playwright need code and break every time a selector changes. SkillForge turns a screen recording of the flow into a SKILL.md an AI agent replays — no selectors to hand-maintain.
Record a test flow — free See QA automationAlready have a Chrome DevTools Recorder export or a Selenium IDE project? Convert it to an agent skill in your browser — free, no signup, nothing uploaded. You can also check an existing SKILL.md the same way.
Selenium and Playwright are powerful, but every test is code, every locator is brittle, and a small UI change can turn a green suite red. Writing and maintaining that suite is a specialized engineering job.
With SkillForge you record the happy-path click-through once. The AI extracts each action into a SKILL.md; an AI agent re-runs it against the app and adapts to the live UI instead of matching hard-coded selectors. Anyone on the team — not just SDETs — can capture a flow.
| Selenium / Playwright | SkillForge | |
|---|---|---|
| Authoring | Write code + selectors | Record the flow |
| Breaks on UI change | Often (selectors) | Agent adapts to the screen — but is not deterministic |
| Cost per run | Free after you write it | Model calls each run |
| Speed | Seconds, parallel across a grid | Slower — an agent reasons step by step |
| Who can author | Engineers | Anyone who can record |
| Best for | Deep CI assertions, edge cases | Fast, resilient happy-path smoke flows |
This is the part a comparison like this usually leaves out, and for a CI audience it matters more than authoring effort. A Selenium suite is expensive to write and free to run. You pay an engineer once, then re-run it a thousand times a week for the cost of the runner.
A recorded skill inverts that. It is free to write and costs something every run, because an agent has to read the screen and decide what to do at each step — model calls, per step, per run, every time. It is slower for the same reason: Selenium fires a click in milliseconds and parallelises across a grid; an agent reasons about the page first.
So the arithmetic is simple. A flow you run on every commit should be code. A flow you run weekly, or one nobody has had time to write at all, is where recording wins — and "nobody has written it" describes far more real test coverage than teams like to admit.
For rigorous assertion-heavy CI pipelines, keep Selenium/Playwright — and note that a recording produces no assertions at all, so what you get is a flow that completes, not a result that is verified (how to add expectations). For flows that survive UI churn, that non-engineers can create, and that run occasionally rather than constantly, recording is faster to author and cheaper to maintain.
Every one of these produces the same artefact: a portable SKILL.md your assistant replays. The format is an open MIT spec, and if you already have a Chrome DevTools Recorder export or a Selenium IDE project you can convert it to a SKILL.md for free without recording anything again. Full index in the guides.
Want to see what you actually get? Download a complete SKILL.md — free, no account, generated from a 34-second screen recording. Hand it to your assistant, or paste it into the validator.
Does SkillForge replace Selenium entirely?
No. It replaces the brittle, high-maintenance happy-path scripts. Keep Selenium/Playwright for deep assertion coverage. Use SkillForge for resilient, quick-to-author flows anyone can record.
How does it handle a changed UI?
Because an AI agent replays the SKILL.md by reading the screen, it adapts to layout changes rather than failing on a hard-coded selector.
How does the cost compare to running a Selenium suite?
It inverts. Selenium is expensive to write and free to run — you pay an engineer once, then re-run it as often as you like for the price of the runner. A recorded skill is free to write and costs model calls on every run, because the agent reads the screen and decides what to do at each step. A flow you run on every commit should be code; a flow you run weekly, or one nobody has written at all, is where recording wins.
Is it as fast as Selenium?
No. Selenium fires a click in milliseconds and parallelises across a grid; an agent reasons about the page before acting, so a replay takes meaningfully longer. If suite runtime is a constraint you are optimising, this is not the tool for that part of your pipeline.
Does it produce assertions?
No, and that is worth knowing before you plan around it. A recording captures what you did, not what should have been true, so nothing identifies which part of the final screen mattered. A passing replay means the flow completed, not that the result was correct. You add an Expected: line to a step yourself — the body of a SKILL.md has no format restrictions, so the agent reads it along with the step.
Do I need Java, Python, or JS?
No. You record the flow; SkillForge writes the SKILL.md. No test code.