Skip to main content

Skills Overview

Skills change how Nimbus approaches a task you describe. They add a specific set of instructions to the planning prompt — like telling a contractor to always use a particular material or follow a specific code standard.

Skills vs Agents​

The key distinction:

  • Skills modify how Nimbus executes a task you already have in mind.
  • Agents replace the task description entirely — the agent knows what to do.
# With a skill: you describe the task, the skill shapes the approach
nimbus run --skill add-tests "write tests for the auth module"

# With an agent: no task description needed
nimbus run --agent test-coverage

Use a skill when:

  • You have a specific file or feature in mind
  • You want Nimbus to apply a particular approach to your own task
  • You're building team-specific conventions into the workflow

Use an agent when:

  • You want a complete, well-defined job done across the whole codebase
  • You don't want to describe the task — you just want the work done

Combining skills and tasks​

Skills compose naturally with any task:

# Adds tests while also following your add-tests conventions
nimbus run --skill add-tests "add tests for the payment processing module"

# Uses your add-logging conventions for a specific file
nimbus run --skill add-logging "add logging to api/routes/webhooks.py"

List skills​

nimbus skills list

Create a skill​

nimbus skills create \
  --name "use-pydantic-v2" \
  --description "Always use Pydantic v2 patterns" \
  --prompt "Use Pydantic v2 model syntax. Never use .dict(), always use .model_dump(). Use model_validator instead of root_validator."