cli-guidelines（CLI guidelines 命令行）

CLI Design Guidelines

Comprehensive guide to designing command-line interfaces following modern best practices. Based on the Command Line Interface Guidelines. Prioritized by impact to guide CLI development and code review.

When to Apply

Reference these guidelines when:

• Creating new CLI tools or adding commands/subcommands
• Implementing help text and error handling
• Parsing arguments and flags
• Designing interactive prompts
• Setting up configuration systems
• Improving CLI UX and output formatting
• Reviewing CLI code for best practices
• Refactoring existing command-line tools

Rule Categories by Priority

Quick Reference

1. The Basics (CRITICAL)

• basics-use-parsing-library - Use argument parsing library (don't roll your own)
• basics-exit-codes - Return 0 on success, non-zero on failure
• basics-stdout-stderr - Send output to stdout, messages/errors to stderr
• basics-help-flags - Support -h and --help flags
• basics-full-flags - Have full-length versions of all flags

2. AI Agent Integration (CRITICAL)

• agents-json-required - Always support --json for agent consumption
• agents-structured-errors - Provide structured error information
• agents-exit-codes-documented - Document all exit codes
• agents-no-prompts-default - Avoid interactive prompts, use flags
• agents-yes-flag - Provide --yes flag to skip confirmations
• agents-progress-to-stderr - Send progress to stderr, data to stdout
• agents-deterministic-output - Ensure deterministic, versioned output
• agents-dry-run - Provide --dry-run for safety
• agents-help-machine-readable - Make help text machine-readable

3. Help & Documentation (CRITICAL)

• help-concise-default - Display concise help when run with no args
• help-lead-examples - Lead with examples in help text
• help-suggest-corrections - Suggest corrections for typos

4. Output Formatting (HIGH)

• output-tty-detection - Check if TTY before using colors/animations
• output-json-flag - Support --json for machine-readable output
• output-plain-flag - Support --plain for script-friendly output
• output-state-changes - Tell the user when you change state
• output-pager - Use a pager for long output

5. Error Handling (HIGH)

• errors-rewrite-for-humans - Catch errors and rewrite for humans
• errors-signal-to-noise - Maintain signal-to-noise ratio in error output
• errors-important-info-end - Put important info at end of output
• errors-exit-code-mapping - Map exit codes to failure modes

6. Arguments & Flags (HIGH)

• args-prefer-flags - Prefer flags over positional arguments
• args-standard-names - Use standard flag names (-f/--force, -n/--dry-run)
• args-no-secrets-flags - Don't read secrets from flags (use files or stdin)
• args-stdin-stdout - Accept - to read from stdin / write to stdout
• args-order-independent - Make flags order-independent

7. Interactivity (HIGH)

• interactive-tty-check - Only prompt if stdin is a TTY
• interactive-no-input-flag - Support --no-input to disable prompts
• interactive-password-no-echo - Don't echo passwords as user types

8. Signals & Control (HIGH)

• signals-exit-on-ctrl-c - Exit immediately on Ctrl-C
• signals-crash-only-design - Design for crash-only operation

9. Robustness (MEDIUM-HIGH)

• robustness-100ms-response - Print something within 100ms
• robustness-progress-indicators - Show progress for long operations
• robustness-validate-early - Validate input early, fail fast
• robustness-idempotent - Make operations idempotent/recoverable
• robustness-network-timeouts - Set timeouts on network operations

10. Subcommands (MEDIUM-HIGH)

• subcommands-consistency - Be consistent across subcommands (same flags, output)
• subcommands-consistent-verbs - Use consistent verbs (create/get/update/delete)
• subcommands-no-abbreviations - Don't allow arbitrary abbreviations
• subcommands-no-catch-all - Don't have catch-all subcommands

11. Configuration (MEDIUM)

• config-precedence - Follow precedence: Flags > Env vars > Project > User > System
• config-xdg-spec - Follow XDG Base Directory spec for config locations

12. Future-proofing (MEDIUM)

• future-additive-changes - Keep changes additive (add flags, don't change behavior)

13. Naming & Distribution (LOW-MEDIUM)

• naming-simple-memorable - Use simple, memorable, lowercase command names
• naming-distribute-single-binary - Distribute as single binary when possible

14. Help & Documentation (MEDIUM)

• help-web-documentation - Provide web-based documentation

15. Analytics (MEDIUM)

• analytics-no-phone-home - Don't phone home without consent

Philosophy

These are the fundamental principles of good CLI design:

1. Human-first design - Design for humans, optimize for machines second
2. Simple parts that work together - Composable via stdin/stdout/stderr, exit codes, JSON
3. Consistency across programs - Follow existing patterns users already know
4. Saying (just) enough - Not too much output, not too little
5. Ease of discovery - Help users learn through suggestions, examples, clear errors
6. Conversation as the norm - CLI interaction is iterative—guide the user through it
7. Robustness - Handle failures gracefully, be responsive, feel solid
8. Empathy - Be on the user's side, help them succeed
9. Chaos - Know when to break the rules—do so with intention

Structure

This skill contains:

• rules/ - Focused, actionable rules

- Each rule has frontmatter (title, impact, tags) - Includes incorrect/correct code examples - TypeScript/JavaScript examples with modern libraries

• references/ - Comprehensive topic guides (for deeper context)

- Background documentation and detailed explanations - Loaded by agents only when deeper understanding is needed

• AGENTS.md - Complete compiled guide (auto-generated from rules)

- All rules in one document - Full table of contents - Searchable reference

How to Use

Read individual rule files for specific guidance:

rules/basics-use-parsing-library.md
rules/agents-json-required.md
rules/output-tty-detection.md

Each rule file contains:

• Brief explanation of why it matters
• Incorrect code example with explanation
• Correct code example with explanation
• Additional context and implementation details

For deeper background, see reference files:

references/philosophy.md
references/basics.md

Full Compiled Document

For the complete guide with all rules expanded: AGENTS.md

Common Argument Parsing Libraries

Further Reading

• Command Line Interface Guidelines
• The Unix Programming Environment
• POSIX Utility Conventions
• 12 Factor CLI Apps