Agent Skills: ansible-patterns

Philosophy

• Keep it dead simple - no mental gymnastics required
• Ansible IS the documentation - if someone can't understand it by reading the playbook, it's too complex
• Explicit and readable over clever and compact

Shell Scripts and Complex Commands

• If a shell command gets complex, create a .sh file instead
• Multi-line shell blocks in tasks are a code smell - extract to script
• Script files are easier to test, debug, and maintain than inline commands
• Reference the script file from the task

Templates and Files

• Always use actual files and Jinja2 templates - reference them from tasks
• Never write file contents directly in task YAML
• Template files go in templates/, static files in files/
• Leverage Jinja2 templating when it makes sense for dynamic content

System Compatibility

• Critical: When writing shell commands, verify the tools/CLI actually exist on target systems
• Don't install missing tools just to make your command work - find a workaround
• Use built-in system tools and common utilities that are guaranteed to be present
• Test against the actual target OS - what works on Ubuntu might not work on CentOS

Task Design

• One task, one purpose - don't chain multiple operations in a single task
• Use descriptive task names that explain what's happening
• Add comments for non-obvious logic
• Use changed_when and failed_when to make tasks idempotent and clear
• Use msg where appropriate with outputs to better understand a task running in front of a user

Anti-Patterns

• Multi-line shell commands in YAML - extract to script if it reaches more than 15-20 lines
• Writing file contents directly in tasks
• Installing tools just to run one command
• Clever one-liners that require decoding
• Tasks that do multiple unrelated things
• Assuming tools exist without verification