atmos-custom-commands - SKILL.md Agent Skill

name: atmos-custom-commands description: "Custom CLI commands: command definition in atmos.yaml, arguments, flags, steps, env vars" metadata: copyright: Copyright Cloud Posse, LLC 2026 version: "1.0.0"

Atmos Custom Commands

Atmos custom commands extend the CLI with project-specific commands defined in atmos.yaml. They appear in atmos help output alongside built-in commands, providing a unified interface for all operational tooling in a project. Custom commands replace scattered bash scripts with a consistent, discoverable CLI.

What Custom Commands Are

Custom commands are user-defined CLI commands configured in the commands section of atmos.yaml. Each command can have:

A name and description
Positional arguments and flags (with shorthand, required/optional, defaults)
Boolean flags
Environment variables (supporting Go templates)
One or more execution steps (shell commands, also supporting Go templates)
Nested subcommands
Access to resolved component configuration via component_config
Authentication via identity
Tool dependencies
Working directory control

Custom commands can call Atmos built-in commands, shell scripts, workflows, or any other CLI tools. They are fully interoperable with Atmos workflows.

Defining Commands in atmos.yaml

Commands are defined under the top-level commands key in atmos.yaml:

# atmos.yaml
commands:
  - name: hello
    description: This command says Hello world
    steps:
      - "echo Hello world!"

Run it with:

atmos hello

Command Structure

Basic Command

commands:
  - name: greet
    description: Greet a user by name
    arguments:
      - name: name
        description: Name to greet
        required: true
        default: "World"
    steps:
      - "echo Hello {{ .Arguments.name }}!"

atmos greet Alice        # Hello Alice!
atmos greet              # Hello World! (uses default)

Command with Flags

commands:
  - name: hello
    description: Say hello with flags
    flags:
      - name: name
        shorthand: n
        description: Name to greet
        required: true
    steps:
      - "echo Hello {{ .Flags.name }}!"

atmos hello --name world
atmos hello -n world

Boolean Flags

commands:
  - name: deploy
    description: Deploy with options
    flags:
      - name: dry-run
        shorthand: d
        description: Perform a dry run
        type: bool
      - name: verbose
        shorthand: v
        description: Enable verbose output
        type: bool
        default: false
      - name: auto-approve
        description: Auto-approve without prompting
        type: bool
        default: true
    steps:
      - |
        {{ if .Flags.dry-run }}
        echo "DRY RUN MODE"
        {{ end }}
        {{ if .Flags.verbose }}
        echo "Verbose output enabled"
        {{ end }}
        {{ if .Flags.auto-approve }}
        terraform apply -auto-approve
        {{ else }}
        terraform apply
        {{ end }}

atmos deploy --dry-run
atmos deploy -d
atmos deploy --auto-approve=false

Boolean flags render as true or false (lowercase strings) in templates.

Flag Defaults

Both string and boolean flags support default values:

flags:
  - name: environment
    description: Target environment
    default: "development"
  - name: force
    type: bool
    description: Force the operation
    default: false

When a flag has a default, users can omit it from the command line.

Trailing Arguments

Arguments after -- are accessible via {{ .TrailingArgs }}:

commands:
  - name: ansible run
    description: Run an Ansible playbook
    arguments:
      - name: playbook
        description: Playbook to run
        default: site.yml
        required: true
    steps:
      - "ansible-playbook {{ .Arguments.playbook }} {{ .TrailingArgs }}"

atmos ansible run -- --limit web
# Runs: ansible-playbook site.yml --limit web

Nested Subcommands

Commands can contain nested subcommands for hierarchical command structures:

commands:
  - name: terraform
    description: Execute terraform commands
    commands:
      - name: provision
        description: Provision terraform components
        arguments:
          - name: component
            description: Component name
        flags:
          - name: stack
            shorthand: s
            description: Stack name
            required: true
        env:
          - key: ATMOS_COMPONENT
            value: "{{ .Arguments.component }}"
          - key: ATMOS_STACK
            value: "{{ .Flags.stack }}"
        steps:
          - atmos terraform plan $ATMOS_COMPONENT -s $ATMOS_STACK
          - atmos terraform apply $ATMOS_COMPONENT -s $ATMOS_STACK

atmos terraform provision vpc -s plat-ue2-dev

Overriding Existing Commands

You can override built-in commands by matching their name:

commands:
  - name: terraform
    description: Execute terraform commands
    commands:
      - name: apply
        description: Apply with auto-approve
        arguments:
          - name: component
            description: Component name
        flags:
          - name: stack
            shorthand: s
            description: Stack name
            required: true
        steps:
          - atmos terraform apply {{ .Arguments.component }} -s {{ .Flags.stack }} -auto-approve

Environment Variables

The env section sets environment variables accessible in steps. Values support Go templates:

commands:
  - name: deploy
    env:
      - key: ATMOS_COMPONENT
        value: "{{ .Arguments.component }}"
      - key: ATMOS_STACK
        value: "{{ .Flags.stack }}"
    steps:
      - atmos terraform plan $ATMOS_COMPONENT -s $ATMOS_STACK

Component Configuration Access

The component_config section resolves the full configuration for a component in a stack, making it available via {{ .ComponentConfig.xxx }} in templates:

commands:
  - name: show-backend
    component_config:
      component: "{{ .Arguments.component }}"
      stack: "{{ .Flags.stack }}"
    steps:
      - 'echo "Backend: {{ .ComponentConfig.backend.bucket }}"'
      - 'echo "Workspace: {{ .ComponentConfig.workspace }}"'

Available fields: .ComponentConfig.component, .ComponentConfig.backend, .ComponentConfig.workspace, .ComponentConfig.vars, .ComponentConfig.settings, .ComponentConfig.env, .ComponentConfig.deps, .ComponentConfig.metadata. For the complete field reference, see references/command-syntax.md.

Go Templates in Steps

Steps support Go template syntax. Access arguments with {{ .Arguments.name }}, flags with {{ .Flags.stack }}, and component config with {{ .ComponentConfig.backend.bucket }}.

steps:
  - "echo Hello {{ .Arguments.name }}"
  - >
    {{ if .Flags.stack }}
    atmos describe stacks --stack {{ .Flags.stack }} --format json
    {{ else }}
    atmos describe stacks --format json
    {{ end }}

Supports if/else, not, eq, and boolean-to-shell conversion. For complete template examples, see references/command-syntax.md.

Authentication

Custom commands can specify an identity for authentication:

commands:
  - name: deploy-infra
    description: Deploy infrastructure with admin privileges
    identity: superadmin
    arguments:
      - name: component
        description: Component to deploy
        required: true
    flags:
      - name: stack
        shorthand: s
        description: Stack to deploy to
        required: true
    steps:
      - atmos terraform plan {{ .Arguments.component }} -s {{ .Flags.stack }}
      - atmos terraform apply {{ .Arguments.component }} -s {{ .Flags.stack }} -auto-approve

The identity applies to all steps. Override at runtime:

# Use command-defined identity
atmos deploy-infra vpc -s plat-ue2-prod

# Override with different identity
atmos deploy-infra vpc -s plat-ue2-prod --identity developer

# Skip authentication
atmos deploy-infra vpc -s plat-ue2-prod --identity ""

The --identity flag is automatically added to all custom commands.

Verbose Output

Control whether step commands are printed before execution:

commands:
  - name: set-eks-cluster
    description: Set EKS cluster context
    verbose: false          # Don't print commands (default is true)
    steps:
      - aws eks update-kubeconfig ...

Working Directory

Control where steps execute:

commands:
  - name: build
    description: Build from repository root
    working_directory: !repo-root .
    steps:
      - make build
      - make test

Path resolution:

Absolute paths used as-is
Relative paths resolved against base_path
!repo-root . resolves to git repository root

Tool Dependencies

Declare tools that must be available before execution:

commands:
  - name: lint
    description: Run tflint on components
    dependencies:
      tools:
        tflint: "0.54.0"
    arguments:
      - name: component
        description: Component to lint
        required: true
    flags:
      - name: stack
        shorthand: s
        description: Stack name
        required: true
    steps:
      - atmos terraform generate varfile {{ .Arguments.component }} -s {{ .Flags.stack }}
      - tflint --chdir=components/terraform/{{ .Arguments.component }}

When you run the command, Atmos:

Checks if the tool is installed at the required version
Installs it from the toolchain registry if missing
Updates PATH to include the tool
Executes the steps

Multiple tools and SemVer constraints are supported:

dependencies:
  tools:
    tflint: "0.54.0"          # Exact version
    checkov: "3.0.0"          # Exact version
    kubectl: "latest"         # Latest available
    terraform: "^1.10.0"      # Compatible range

Common Patterns

Common custom command patterns include: listing stacks/components, setting EKS cluster context, security scanning, cost estimation, and documentation generation. For complete examples of each, see references/command-syntax.md.

Quick Example: List Stacks

commands:
  - name: list
    description: List stacks and components
    commands:
      - name: stacks
        description: List all Atmos stacks
        steps:
          - >
            atmos describe stacks --process-templates=false --sections none | grep -e "^\S" | sed s/://g

Quick Example: Security Scan with Dependencies

commands:
  - name: security-scan
    description: Run security scans on infrastructure code
    dependencies:
      tools:
        tflint: "0.54.0"
        checkov: "3.0.0"
    steps:
      - tflint --chdir=components/terraform
      - checkov -d components/terraform

Best Practices

Provide descriptive names and descriptions. Commands show in atmos help, so clear descriptions help team members discover and understand available tooling.
Use Go templates for conditional logic. Rather than writing separate commands, use template conditionals to handle optional flags.
Set sensible defaults. Use the default attribute on arguments and flags so common cases require minimal input.
Use component_config for context-aware commands. When a command needs to know about a component's resolved configuration (backend, vars, workspace), use component_config instead of hardcoding values.
Use verbose: false for noisy commands. Suppress command echo for commands that produce a lot of output or contain sensitive information.
Leverage tool dependencies. Instead of documenting prerequisites, declare them in dependencies.tools so they are auto-installed.
Organize with nested subcommands. Use nested commands for related operations (e.g., atmos list stacks, atmos list components).
Combine with workflows. Use custom commands for atomic operations and workflows for multi-step orchestration. They can call each other.
Use !repo-root . for working_directory when commands need to run from the repository root regardless of where atmos is invoked.
Use environment variables for shared values. Define values once in env and reference them in multiple steps via shell variables like $ATMOS_COMPONENT.

Additional Resources

For the complete custom command YAML syntax reference, see references/command-syntax.md