cli-design（CLI 设计）

CLI Design: Unix-Composable Command-Line Interfaces

This skill covers language-agnostic CLI design principles. The rules about stream separation, exit codes, format flags, and composability apply regardless of implementation language.

For API contract stability and Hyrum's Law, see the api-design skill. For config, env vars, and graceful shutdown, see the twelve-factor skill.

TypeScript implementation patterns are in the resources/ directory. Load them on demand when building a CLI in TypeScript:

When to Use

• Building any command-line tool (any language)
• Designing command tree, flags, and I/O contracts
• Implementing the output layer (format detection, stream routing)
• Testing CLI behavior (stdout/stderr separation, exit codes)
• Reviewing a CLI for Unix composability

Core Principle

stdout is for DATA only — the product the user asked for. stderr is for EVERYTHING ELSE — diagnostics, progress, spinners, warnings, errors.

This separation is what makes mycli --json | jq ... work. One spinner character on stdout breaks every downstream pipe.

"Whatever software you're building, you can be absolutely certain that people will use it in ways you didn't anticipate. Your software will become a part in a larger system — your only choice is over whether it will be a well-behaved part." — clig.dev

The Unix Stream Contract

Buffering behavior:

• stdout: line-buffered when connected to a TTY, block-buffered when piped (~2x faster than stderr)
• stderr: unbuffered — every write is a syscall (immediate but expensive)
• Check each stream independently — stdout being piped does not mean stderr is piped

When stdout is piped, the user doesn't want your status messages in their data. All non-data output must go to stderr.

For a deep dive on buffering behavior and performance implications, see resources/stream-contracts.md.

Keep Handlers Pure

The practical rule: functions that do the work should return data, not write to stdout. The CLI entry point handles all I/O.

Entry point (CLI main)              Your logic (handlers)
─────────────────────               ─────────────────────
parse args                          (input) → structured result
detect format (json/plain/human)    no printing to stdout
call handler                        no writing to stderr
format the result                   no calling exit
write to correct stream             just returns data
set exit code

This isn't an architecture mandate — it's just clean function design. The benefits are concrete:

• Testable without subprocess spawning — call the handler, assert on the returned value
• Format flexibility for free — same data renders as JSON, plain text, or coloured tables by swapping one function
• Reusable — the same handler works from a CLI, MCP server, HTTP API, or programmatic import

For simple CLIs where the "handler" is just calling a library, this separation already exists naturally — your library returns data, your CLI formats it. No extra layers needed.

If your project uses hexagonal architecture, the mapping is direct: the CLI entry point is a driving adapter, and the handler is a use case that returns a result through a port. See the hexagonal-architecture skill — the patterns reinforce each other, but hex arch is not required to benefit from keeping handlers pure.

For TypeScript implementation patterns (Result types, entry point wiring, formatters, logger interfaces), see resources/output-architecture.md.

Format Flag Contract

Three-tier output hierarchy:

Default: Human-Readable

• Colors, tables, formatted text
• Progress bars and spinners on stderr
• Output tailored for terminal width
• May change between versions — this is not a contract

--plain: Grep/Awk-Friendly

• One record per line, no formatting, no colors
• Stable between minor versions — this is a contract
• Flat table rows, no borders, no grouped sections
• Enables: mycli list --plain | grep error | wc -l

"Encourage your users to use --plain or --json in scripts to keep output stable." — clig.dev

--json: Structured Data

• stdout contains ONLY valid JSON — no spinners, no color, no progress
• stderr continues normally — human diagnostics still visible
• Errors are structured JSON too — not just success responses
• Schema is versioned — breaking changes to JSON output are breaking changes to the CLI
• --json implies non-interactive regardless of TTY

Consistent envelope:

{ "ok": true, "data": { ... } }
{ "ok": false, "error": { "code": "CONFIG_MISSING", "message": "...", "fix": "..." } }

NDJSON for Streaming

For large datasets, use NDJSON (one JSON object per \n):

• Each line is independently parseable
• Include a type field per record for multiplexing events
• Final line can be a summary record
• Enables: mycli run --format ndjson | while read -r line; do ...; done

For NDJSON specification details, see resources/stream-contracts.md.

Exit Codes

Rules:

• Non-zero exit code MUST have a stderr explanation
• Document exit codes in --help
• Never use codes above 125 for application errors (reserved for signals: 128 + signal number)
• Exit code 75 (transient) is critical — it tells retry logic the failure may be temporary
• Map non-zero codes to the most important failure modes for your tool

TTY Detection

Check priority order (first match wins):

Check stdout and stderr independently. When stdout is piped but stderr is a TTY, you can still show spinners on stderr while keeping stdout clean for the pipe consumer.

Optionally support MYCLI_NO_COLOR for app-specific color override.

Input Design

Flags Over Arguments

• 1 positional arg: acceptable (the "main thing")
• 2 positional args: suspicious — consider flags instead
• 3+ positional args: never acceptable

Flags are self-documenting, order-independent, and future-proof.

# Bad — which is source, which is destination?
mycli copy myapp backup

# Good — explicit
mycli copy --from myapp --to backup

Standard Flags

Always provide long forms. Short flags only for the most common operations.

Prompts and Interactivity

• All prompts MUST be bypassable via flags for scriptability
• Confirmation → --yes or --force
• Selection → --type=value
• Text input → --name=value
• Passwords → --password-file=path or stdin pipe
• If stdin is not a TTY, never prompt — fail with a clear error or use defaults
• Secrets via files/stdin/env only — never via flag values (they leak to ps output and shell history)

Conventions

• Support -- to stop flag parsing: mycli run -- --flag-for-child-process
• Support - for stdin/stdout file arguments: curl ... | mycli process -
• Accept both --flag=value and --flag value
• If stdin is expected but is an interactive terminal, display help immediately (don't hang like cat)

Config Precedence

Highest to lowest priority:

1. Flags — per-invocation overrides
2. Environment variables — MYCLI_* prefix, per-session
3. Project config — .myclirc, mycli.config.ts, or in package.json
4. User config — ~/.config/mycli/ (follow XDG spec)
5. Defaults — sensible built-in values

Rules:

• Follow the XDG Base Directory Specification for config file locations
• Env var naming: MYCLI_* prefix, uppercase letters + digits + underscores
• Never accept secrets via flags — use env vars, files, or stdin
• Read .env where appropriate, but don't use it as a substitute for proper config
• If you modify configuration that belongs to another program, ask consent first

Error Design

Every error needs:

1. Machine-readable code — UPPER_SNAKE_CASE (e.g. CONFIG_MISSING, AUTH_EXPIRED)
2. What went wrong — context: which resource, operation, input
3. How to fix it — exact command or action the user should take
4. Reference — docs URL or mycli help <topic> (optional)

Human Mode

Error: CONFIG_MISSING — Configuration file not found
No configuration file found at ./mycli.config.ts or ~/.config/mycli/config.ts

Fix: Run `mycli init` to create a default configuration file
Docs: https://mycli.dev/docs/configuration

• Put the most important information last (the eye is drawn to the end)
• Use red sparingly and intentionally
• Suggest corrections for typos ("Did you mean 'deploy'?")
• Group similar errors under one header — don't repeat 50 similar-looking lines
• Write debug logs to a file, not the terminal (unless --debug)

JSON Mode

Errors are structured too — not just success responses:

{
  "ok": false,
  "error": {
    "code": "CONFIG_MISSING",
    "message": "No configuration file found at ./mycli.config.ts",
    "fix": "Run `mycli init` to create a default configuration file",
    "transient": false
  }
}

The transient boolean tells retry logic whether the failure may be temporary.

Composability Patterns

Design for real-world pipes:

# Filter structured output
mycli list --json | jq '.data[] | select(.status == "failed")'

# Stream results for large datasets
mycli run --format ndjson | while read -r line; do echo "$line" | jq '.file'; done

# Feed stdin
cat previous-results.json | mycli report --format markdown

# Combine with other tools
mycli run --json | mycli diff --baseline previous.json

# Silent mode for CI — only exit code matters
mycli check --quiet || echo "Check failed!"

# Chain: create outputs an identifier, next command uses it
mycli create --json | jq -r '.data.id' | xargs mycli deploy --id

# Column selection for efficiency
mycli list --json --fields name,status,id | jq '.data[]'

# Parallel processing
mycli list --json --fields id | jq -r '.data[].id' | xargs -P4 mycli process --id

Key patterns:

• Create commands output identifiers so subsequent commands can chain
• List commands support --fields for column selection (reduces output size, critical for agent efficiency)
• --quiet for CI scripts that only care about the exit code
• NDJSON for streaming large datasets without buffering everything in memory
• --dry-run with --json outputs planned changes as structured data

Subcommand Design

• noun verb pattern is most common: mycli config set, mycli report generate
• Be consistent across all subcommands — same flag names for same things
• No ambiguous pairs (update vs upgrade is confusing)
• No catch-all subcommands (you can never add subcommands with conflicting names)
• No arbitrary abbreviations — aliases must be explicit and stable
• With no args: list subcommands (multi-command CLI) or show help (single-command CLI)

Help

• mycli --help — top-level help
• mycli help <subcommand> — subcommand help
• mycli <subcommand> --help — same as above
• If run with missing required args, show concise help + 1-2 examples + "use --help for more"
• Examples are the most-read section — lead with them
• Include flag types, defaults, and allowed values for finite sets

Output Stability Contract

Stdout is a public API. Breaking changes to stdout format are breaking changes to the CLI.

When in doubt, add alongside — don't modify. Deprecate with stderr warnings before removing.

Anti-Patterns

Verification Checklist

After designing or reviewing a CLI:

• [ ] stdout has ONLY data; stderr has everything else
• [ ] Every command supports --json with consistent envelope
• [ ] Exit codes are semantic and documented in --help
• [ ] Every prompt has a --yes/--force/--flag bypass
• [ ] Errors include: code, message, fix suggestion
• [ ] --dry-run available for mutating commands
• [ ] Progress/spinners go to stderr, never stdout
• [ ] NO_COLOR, TERM=dumb, and --no-color respected
• [ ] Piped output contains zero ANSI escape codes
• [ ] Success output includes next-action guidance
• [ ] Existing flags, exit codes, output fields never removed or renamed
• [ ] JSON schema is versioned (additions safe, removals breaking)
• [ ] Config follows flags > env > project > user > defaults
• [ ] Secrets accepted only via files/stdin/env, never via flags
• [ ] Startup < 500ms, print something in < 100ms
• [ ] Ctrl-C exits fast with bounded cleanup
• [ ] --help includes 2-3 realistic examples
• [ ] Human output is grep-parseable (flat rows, no table borders)

Quick Reference

Stream Routing

stdout ← data, results, JSON, NDJSON
stderr ← progress, spinners, warnings, errors, debug, prompts

Format Hierarchy

Default (TTY)     → colors, tables, formatted text
--plain           → one record per line, stable, grep-friendly
--json            → structured JSON, versioned schema
--format ndjson   → streaming, one JSON object per line

Exit Codes

0   success
1   domain failure
2   invalid usage
75  temporary failure (retry)
78  config error
130 SIGINT (Ctrl-C)
143 SIGTERM

Config Precedence

flags > env vars > project config > user config > defaults

Flags Cheat Sheet

-h  --help        Show help
    --version     Print version
-q  --quiet       Less output
-v  --verbose     More output
-d  --debug       Diagnostic output
-f  --force       Skip prompts
-n  --dry-run     Preview changes
    --json        Structured JSON
    --plain       Grep-friendly text
    --no-color    Disable color
    --no-input    No prompts
-o  --output      Output file
    --fields      Select columns