Skills are know-how. A workflow is the process.
You don’t choose between them. Skills hold the expertise: your own, your team’s, or a library you installed. A workflow sets the steps, suggests or requires a skill at each one, and sets who signs off.
| Question | A skill | A Forge workflow |
|---|---|---|
| Who decides what happens next | Your AI, following the instructions | Forge, one step at a time |
| A person signing off | Only if the instructions ask for it | Set for each step: Never asks, AI decides, or Always asks |
| Sharing it with a team | Files in a repo, or a plugin | Set once by an admin, for the organization or one team |
| What’s left afterward | The chat transcript | A record of each step |
The right skill, at the right step.
Each time a workflow runs, Forge sees the skills your AI tool already has: your own, your team’s, and any library you installed. At each step it names the ones that fit, such as a brainstorming skill while you scope a feature, or a review skill while you check a branch. You decide whether to use them.
Or required, when it matters.
Some steps shouldn’t be optional. In the Forge dashboard’s workflow editor, an admin can mark a skill as required for a step. Forge asks the AI to use that skill before the step completes, every time the step runs.
The Forge workflow editor, with a brainstorming skill from the Superpowers library required on the “Gather Context” step.