LifeOS
← All 17 principles

Principle № 8 · Architecture

CLI as Interface

Command-line first, with keyboard shortcuts and automation

CLI as Interface

Overview

Every operation should be accessible via command line. If there’s no CLI command for it, you can’t script it or test it reliably.

Command-line tools are faster, more scriptable, and more reliable than graphical interfaces. GUIs are great for discovery and exploration. CLI is essential for automation, testing, and production use.

The pattern: build the CLI first. GUIs can wrap CLI commands. But if there’s no underlying CLI, automation becomes impossible.

Why This Matters

CLI is automatable - GUI clicks can’t be scripted reliably. CLI commands compose into workflows, get version-controlled, and run in CI/CD pipelines.

CLI is testable - You can’t write automated tests for button clicks. You can test CLI commands exhaustively.

CLI is fast - Expert users operate at keyboard speed. GUI users operate at mouse speed. For frequent operations, this compounds.

CLI is reproducible - “Click the third icon, then the dropdown, then…” vs. skill-name workflow --option=value. One is reproducible, one is not.

CLI is portable - Commands work over SSH, in Docker containers, in CI/CD. GUIs require graphics stacks and human interaction.

CLI is discoverable - command --help shows all options. GUIs hide functionality behind menus and modals.

Implementation

LifeOS is CLI-first throughout:

Every skill exposes CLI commands - Skills in .claude/Skills/ have tools/ directories with executable scripts. Text in, text out, compose via pipes.

Skill routing defines commands - SKILL.md in each skill documents available commands, arguments, and usage patterns.

Keyboard shortcuts - Claude Code supports /skillname shortcuts. Type /research instead of clicking through menus.

Hook system is CLI-based - Hooks are TypeScript scripts that run on events. Everything scriptable, nothing manual.

Workflows are command sequences - Research workflows compose CLI tools. extract-content | fabric/extract_wisdom | format-output

MCP integration - Model Context Protocol provides CLI-like tool interfaces. Standardized, composable, testable.

Examples

Example 1: Research Workflow GUI way:

  1. Open browser
  2. Navigate to research skill interface
  3. Click “New Research”
  4. Fill form with topic
  5. Click “Start”
  6. Wait for results
  7. Click “Export”
  8. Choose format
  9. Save file

CLI way:

research deep-dive "AI infrastructure patterns 2024-2025" --sources=5 --format=markdown > output.md

One command. Scriptable. Reproducible. Testable.

Example 2: Content Generation GUI way:

  • Open Art skill
  • Select diagram type
  • Enter description
  • Click generate
  • Wait
  • Download
  • Optimize separately
  • Upload manually

CLI way:

art generate-diagram "LifeOS architecture" --type=technical --optimize --output=public/images/

Example 3: Skill Management GUI way:

  • Navigate to skills directory
  • Create folders manually
  • Copy template files
  • Edit configurations
  • Test manually
  • Document separately

CLI way:

createskill --name=MySkill --type=research --template=standard