agentleFS
joshuadavidthomasjoshuadavidthomas/agent-skills/writing-cli-skills/SKILL.md

writing-cli-skills

Skill49 starsChanged 6 months agoMITView on GitHub

Use when authoring an agent skill that wraps a command-line tool — covers hands-on tool exploration, required vs. recommended sections, installation/usage structure, trigger-rich descriptions, task-grouped commands, progressive disclosure, and a pre-publish checklist. Triggers for CLI / command-line / terminal / shell-command tools and binary wrappers; for review, run the Checklist section.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 6 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run --help on every subcommand

    line 13
  2. Distinctive:

    Delete sections that don't apply

    line 18
  3. Very specific:

    Do not dump --help output verbatim — summarize the useful parts

    line 43
  4. Distinctive:

    Do not document every flag — focus on the 80% use cases

    line 44
  5. Distinctive:

    Do not include commands you haven't tested

    line 45
  6. Distinctive:

    Do not exceed 500 lines — this bloats agent context windows

    line 46
  7. Distinctive:

    Every CLI skill needs at minimum:

    line 52
  8. Very specific:

    Keep SKILL.md under ~500 lines.

    line 101
  9. Distinctive:

    Before publishing a CLI skill:

    line 124

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Writing CLI Skills

How to write an agent skill for a command-line tool.

Quick Start

  1. Install the tool and play with it — don't just read docs. If the tool is unavailable, use WebFetch on official docs + man pages, but note this in your skill as "not hands-on verified"
  2. Run --help on every subcommand
  3. Try the common operations yourself
  4. Note what surprises you or isn't obvious
  5. Copy references/template.md to your new skill directory
  6. Fill in sections based on your hands-on experience
  7. Delete sections that don't apply
# First: actually use the tool
my-tool --help
my-tool subcommand --help
my-tool do-something  # try it!

# Then: create the skill
# This will depend on the tool

# Claude Code
ln -s "$PWD/skills/my-tool" ~/.claude/skills/my-tool

# OpenCode
ln -s "$PWD/skills/my-tool" ~/.config/opencode/skills/my-tool

# Clawdbot
ln -s "$PWD/skills/my-tool" ~/clawd/skills/my-tool

Reading docs is no substitute for hands-on use. You'll discover defaults, gotchas, and real behavior that docs don't mention.

What NOT to Do

  • Do not dump --help output verbatim — summarize the useful parts
  • Do not document every flag — focus on the 80% use cases
  • Do not include commands you haven't tested
  • Do not exceed 500 lines — this bloats agent context windows

Sections

Required

Every CLI skill needs at minimum:

SectionPurpose
Frontmattername, description (with trigger phrases)
InstallationHow to get the binary
UsageThe 80% use cases

Recommended

SectionWhen to Include
RequirementsTool needs accounts, API keys, or dependencies
Quick StartTool has a simple "happy path"
Output FormatsTool can output JSON or custom formats
Tips & GotchasTool has some edge cases or things an agentic LLM should not use
TroubleshootingTool has debug modes or common failure modes
ConfigurationTool has config files or env vars
UninstallTool leaves config/data behind
ReferencesWhen there are useful docs or content that contains more details

Writing Good Descriptions

Include trigger phrases so the agent knows when to load the skill:

# Good
description: Monitor RSS feeds for updates. Use when tracking blogs, checking for new posts, or building a feed reader workflow.

# Bad  
description: RSS tool.

Organizing Commands

Group by task, not by command name:

## Usage

### View / List
### Create / Add  
### Edit / Update
### Delete / Remove
### Search

Progressive Disclosure

Keep SKILL.md under ~500 lines. Move details to references/:

my-tool/
├── SKILL.md                    # Core usage
├── references/
│   ├── advanced-config.md      # Deep config docs
│   └── api-reference.md        # API details
└── scripts/
    └── helper.sh               # Helper scripts

Frontmatter Reference

---
name: tool-name                  # Required, matches directory
description: What + when         # Required, include triggers
---

Checklist

Before publishing a CLI skill:

  • Frontmatter has name + description with trigger phrases
  • Installation covers target platforms
  • Includes verification command (tool --version)
  • Config file locations documented
  • Required env vars listed
  • Common operations in usage cover 80% of use cases
  • Examples show realistic usage with sample output
  • Output formats documented (if tool supports JSON/etc)
  • Troubleshooting includes debug mode
  • Uninstall cleans up config and data
  • Under 500 lines (details in references/)

Template

See references/template.md for a complete starting point.

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.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

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.