Build an agent from a skill
2 min · 5 steps
New Agent → From an agent skill → pick the folder → the Build model writes the description → review and build, just like a typed prompt.
Skill-to-agent lives inside the New Agent dialog as the second option, 'From an agent skill'. Pick a skill folder (anything with a SKILL.md at its root); your Build model reads it and writes the agent description. You land on the prompt step with that description pre-filled — review or tweak it, then build the agent exactly as you would from a typed prompt. It needs a Build model configured (Settings → AI → AI Model for Build), since the analysis runs there.
Steps
- Open New Agent.
Sidebar → Agents → New Agent. The dialog opens on a chooser with three options: Prompt with Cloud AI Model, From an agent skill, and Build Manually.
- Choose 'From an agent skill'.
Click the middle card. (If it's greyed out, you haven't set a Build model yet — the card links you to Settings → AI → Cloud Models Config.)
- Pick the skill folder.
Click 'Choose skill folder…' and select the directory containing SKILL.md. Common locations: ~/.claude/skills/<name>/ (a personal Claude Code skill) or any `git clone`'d skill repo.
- Analyze & Continue.
Avery's Build model reads the skill and writes the agent description. You land on the prompt step with that description pre-filled — read it over and edit anything you'd phrase differently.
- Generate the agent.
Click 'Generate Agent'. From here it's identical to building from a typed prompt: the model designs the graph, you review it, and you land in the Builder. Edit nodes, change the trigger, run it.
Live recipes need the desktop
This article is a static preview. The in-app Help sidecar inside Avery NXR can fire each step against your live project — install the desktop to use it interactively.