agentleFS
notquenotque/vexjoy-agent/skills/engineering/cli-design/SKILL.md

cli-design

Skill448 starsChanged 4 days agoMITView on GitHub

Design a CLI interface: args, flags, help, output, errors, exit codes, config.

17 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Gate: every flag used in the examples appears in the flags table, and every failure mode shown maps to an exit code.

    line 69
  2. Distinctive:

    Keep the spec language-agnostic.

    line 74

“Common” is measured: kept by at least 6 in 10 of the 13 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

CLI Design

Design a command-line tool's interface before implementation: human-first, script-friendly, Linux-only. Output is a compact spec the user or an agent can implement directly. Rubric source: clig.dev (rebuilt as references/clig-checklist.md).

Deep References

WhenLoadContent
Phase 2: applying design rubricreferences/clig-checklist.mdCLI design checklist condensed from clig.dev

Workflow

Phase 1: SCOPE

Lock the interface with the minimum questions. Proceed with the conventions in Phase 2 when the user is unsure.

  • Command name and one-sentence purpose.
  • Primary user: humans, scripts, or both.
  • Input sources: args vs stdin; files vs URLs. Secrets travel via file or stdin, because flags leak through ps and shell history.
  • Output contract: human text, --json, --plain, exit codes.
  • Interactivity: prompts allowed? --no-input needed? confirmation for destructive ops?
  • Config model: flags, env, config file; precedence.

Gate: name, purpose, and I/O contract are known. Proceed only when gate passes.

Phase 2: DESIGN

Load references/clig-checklist.md and apply it as the default rubric. For each section, pick the convention and record it in the spec. Diverge from a convention only deliberately, and document the divergence in the spec — interfaces are contracts, and surprising contracts break scripts.

Phase 3: DELIVER

Produce the spec from this skeleton. Drop a section only when it genuinely has no content; fill every other section.

  1. Name and one-liner: command name plus a single sentence of purpose
  2. Usage line: the synopsis as --help will print it, global flags and subcommand slot included
  3. Subcommands: purpose of each, whether it mutates state, whether re-running it is safe
  4. Args/flags table: columns for name, type, default, required?, example
  5. I/O contract: primary data and machine-readable output on stdout; everything else (errors, progress, logs) on stderr
  6. Exit codes: map each failure mode to a code — success 0, failure 1, bad usage 2; mint extra codes only for cases scripts must distinguish
  7. Safety: --dry-run, confirmation rules, --force, --no-input
  8. Env/config: env vars; config file path; precedence order with flags highest, then env, project config, user config, system
  9. Examples: enough invocations to cover the common flows; show at least one pipeline or stdin use

Gate: every flag used in the examples appears in the flags table, and every failure mode shown maps to an exit code.

Constraints

  • Stay at spec altitude: when the request is "design the interface," deliver the spec and stop. Implementation is a separate task.
  • Keep the spec language-agnostic. Recommend a parsing library only when asked.
  • Target Linux. Skip Windows/macOS path, signal, and packaging concerns.

Error handling

Request mixes design and implementation

Cause: user says "design and build." Solution: deliver the spec first, get confirmation, then implement against it.

Spec balloons past one page

Cause: subcommand sprawl or speculative flags. Solution: cut flags that lack a named user need; defaults should serve most users without aliases.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.