What if teaching an AI agent was as simple as recording your screen?
No coding. No brittle selectors. No debugging when the UI changes. Just perform the task once, and the AI learns how to do it.
That's the promise of SkillForge and the SKILL.md format.
The Old Way: Writing Code
Traditional browser automation requires you to:
- Inspect HTML elements
- Write CSS selectors or XPath expressions
- Handle timing issues and wait conditions
- Update scripts every time the website changes
This approach is fragile. A button moves, a class name changes, and your automation breaks. You're not describing what you want done—you're prescribing exactly how to do it.
The New Way: Recording Intent
SkillForge flips this model. Instead of writing code, you simply record yourself performing a task. The AI analyzes the recording and extracts:
Goals: What you're trying to accomplish
Workflows: The sequence of steps
Context: How to identify UI elements semantically
Decisions: Where choices are made along the way
The output is a SKILL.md file—a structured, human-readable document that any compatible AI agent can execute.
Example SKILL.md
Here's what a generated skill looks like:
# Submit Expense Report
## Goal
Submit an expense report in the company's HR portal
## Workflow
1. Navigate to hr.company.com/expenses
2. Click "New Expense Report"
3. Fill in date, amount, and category
4. Upload receipt image
5. Submit for approval
## Context
- "New Expense Report" button: Primary CTA on dashboard
- Date field: Date picker labeled "Expense Date"
- Amount field: Number input labeled "Amount ($)"
- Category dropdown: Select labeled "Category"
Notice what's missing: no CSS selectors, no XPath, no brittle identifiers. Just semantic descriptions that an AI agent can interpret.
Why This Matters
For Developers: No more maintaining broken automation scripts. Skills survive UI updates because they describe intent, not implementation.
For Domain Experts: Create automation without learning to code. If you can perform a task, you can teach an AI agent to do it.
For Teams: Build reusable skill libraries. Document processes through demonstration rather than documentation that becomes outdated immediately.
Real-World Applications
- Sales teams: Record your CRM workflow once, automate it forever
- Operations: Document internal processes through demonstration
- Customer support: Create automated troubleshooting flows
- Data entry: Eliminate repetitive form filling
- QA testing: Generate browser automation tests from user sessions
The Future of Agent Training
We're moving toward a world where AI agents learn by observation, just like humans. The SKILL.md format is a step toward standardizing how we capture and share agent capabilities.
Get started with SkillForge:
🔗 https://www.producthunt.com/products/skillforge-2
What workflows will you teach your AI agents?
Top comments (0)