If you've ever built browser automation, you know the pain. You spend hours crafting the perfect script, testing every selector, handling edge cases—and then the website updates.
Suddenly your automation breaks. Buttons moved. CSS classes changed. Your carefully constructed house of cards collapses.
There's a better way.
The Root Problem
Traditional automation describes how to do things, not what needs to be done.
When you write:
await page.click('#submit-btn');
You're not saying "submit the form." You're saying "click the element with ID submit-btn." When that ID changes, your script breaks—even though the goal (submitting the form) hasn't changed.
Intent Over Implementation
What if instead of prescribing exact steps, you could describe the goal and let the AI figure out the details?
That's the approach behind SkillForge—a tool that captures intent rather than mechanical implementation.
Record: Perform any web task while recording your screen.
Extract: AI analyzes the recording to understand your goal, the workflow, and the context.
Generate: The output is a SKILL.md file—a structured, human-readable description of the task.
Execute: The agent follows the skill description, adapting to UI changes because it understands what it's trying to accomplish.
Why This Changes Everything
This shift from "how" to "what" changes everything:
- Skills survive UI updates because they describe intent
- Domain experts can create automation without coding
- Skills are portable across frameworks (AutoGen, LangChain, CrewAI)
- The format is auditable and version-controllable
Get Started
SkillForge is live on Product Hunt today:
🔗 https://www.producthunt.com/products/skillforge-2
What workflows would you automate if brittle selectors weren't a concern?
Top comments (0)