claude-skills-templates

Five production-ready Claude Skills you can drop into Claude Code, Claude.ai, or the Claude API. Each skill demonstrates a different SKILL.md pattern: pure-prompt, structured workflow, configurable reference file, executable script, and decision-support.

Skills follow the Agent Skills open standard. Each one is a folder with a SKILL.md file plus optional supporting resources.

Skills index

Skill Pattern What it does
one-pager Pure prompt Turns notes or transcripts into a structured one-page brief
code-reviewer Structured workflow Five-stage code review (architecture, quality, tests, performance, security)
brand-voice Configurable reference Applies a tone and voice guide to drafts; user fills in their voice rules
csv-cleaner Skill plus script Diagnoses and cleans messy CSV or TSV files
decision-doc Decision support Structured tradeoff analysis with explicit flip conditions

How skills work

Skills use progressive disclosure. Three loading levels:

  1. Metadata (name and description in YAML frontmatter): always visible to Claude, used to decide whether to load the skill.
  2. SKILL.md body: loaded into context only when the skill triggers.
  3. Bundled resources (scripts, references, assets): loaded on demand.

The description field is the primary triggering mechanism. Write it to be specific about both what the skill does AND when to use it.

Installation

Claude Code (CLI)

Copy the skill folder into your project's .claude/skills/ directory or your global ~/.claude/skills/ directory:

cp -r skills/one-pager ~/.claude/skills/

Claude Code picks it up on the next session.

Claude.ai

Zip the skill folder and upload it via the Skills menu:

cd skills/one-pager && zip -r ../one-pager.skill .

Then upload the resulting .skill file in Claude.ai under Settings, Skills.

API

Pass the skill via the skills parameter when calling the Messages API. See Anthropic's Skills documentation for the current syntax and supported runtime.

Customizing

Every skill in this repo is a starting point. Recommended workflow:

  1. Copy the skill folder into your project.
  2. Open SKILL.md and edit the description field to match your trigger phrases.
  3. Edit the body to match your team's conventions.
  4. For brand-voice, fill in references/brand-voice.md with your actual voice rules.
  5. Test with 5 to 10 sample prompts before deploying.

Description writing tips

The description field is a single string in the YAML frontmatter. Two failure modes to avoid:

  • Undertriggering: description is too generic, Claude does not load the skill when it should. Fix: name specific trigger phrases users actually say.
  • Overtriggering: description is too broad, Claude loads the skill for unrelated requests. Fix: name the contexts where the skill does NOT apply.

A good description includes:

  • One sentence on what the skill does
  • A list of trigger phrases ("use when the user says X, Y, Z")
  • One sentence on a guardrail ("always read references/X.md first" or "do not use for ad-hoc summarizing")

Contributing

Pull requests welcome. New skills should:

  • Have a single, specific purpose, not "do everything"
  • Include a description that triggers reliably without overtriggering
  • Stay under 500 lines in SKILL.md (use bundled resources for longer content)
  • Include 2 to 3 example trigger phrases in the description

More

Part of a catalog of single-file browser tools and plain-language references, all MIT licensed and dependency-free: 0xelitesystem.github.io. Built by elitesystem.ai.

License

MIT. See LICENSE.