I built SkillForge because every team seems to have the same graveyard: a folder full of onboarding videos, Looms, and walkthroughs that nobody turns into reusable systems.
A founder records a demo. An ops lead records the "right way" to update a dashboard. A teammate explains a messy browser workflow once. Then the video gets watched maybe twice, forgotten, and six weeks later someone asks for the exact same process again.
That felt broken to me.
I didn't want another documentation tool. I wanted a way to turn a screen recording into something executable: a reusable AI agent automation that can actually replay the workflow instead of just describing it.
That's why I built SkillForge.
The problem with traditional SOPs
Most SOPs fail for a simple reason: they decay immediately.
- Videos are hard to search
- Written docs drift from the actual UI
- New hires still need hand-holding
- Browser workflows stay trapped in someone's memory
The result is that companies keep re-teaching the same task instead of operationalizing it.
What I wanted instead
My goal was simple:
- Record a workflow once
- Let AI extract the real sequence of actions
- Turn it into a reusable automation another agent can run again
No code. No giant setup process. No forcing non-technical people to think like automation engineers.
If someone can show a process on screen, they should be able to turn that process into an asset.
Why this matters for teams
I think the most interesting part of AI automation is not "one more chatbot."
It's converting messy human know-how into repeatable systems.
For example:
- onboarding a VA to do repetitive browser work
- preserving founder knowledge before it becomes tribal knowledge
- turning one-off support workflows into repeatable playbooks
- capturing brittle internal processes before the person who knows them leaves
That is the gap I built SkillForge for.
The product philosophy
A lot of automation products start from the wrong place. They start from flows, nodes, prompts, and configuration.
I wanted to start from demonstration.
Humans are much better at showing than specifying. Recording a workflow is natural. Designing a robust automation from scratch is not.
So SkillForge starts with the thing people already know how to do: show the task.
If you want to try it, here it is:
https://skillforge.expert?utm_source=devto&utm_medium=social&utm_campaign=launch_feb2026
I'm still early, still iterating, and still obsessed with closing the gap between "I can do this task" and "my AI agent can do this task too."
If that problem is interesting to you, I'd love to hear how you're thinking about turning demos, SOPs, and walkthroughs into actual automation.
Top comments (0)