Skill Creator

Skill Creator

A guide for creating well-structured, reusable Claude Code skills.

Communicating with the user

Pay attention to context cues to gauge familiarity. Keep explanations accessible — don't assume deep technical knowledge unless the user demonstrates it.

---

Creating a skill

Capture Intent

Understand what the skill should enable. Extract from the current task context:

1. What should this skill enable Claude to do?
2. When should this skill trigger? (what user phrases / contexts)
3. What's the expected output format or artifact?

Interview and Research

Proactively think about edge cases, input/output formats, success criteria, and dependencies before writing. If useful libraries or APIs are involved, check what's available in the environment first.

Write the SKILL.md

Based on the task requirements, fill in these components:

• name: Skill identifier (must match the folder name)
• description: When to trigger, what it does. This is the primary triggering mechanism — include both what the skill does AND specific contexts for when to use it. To combat undertriggering, make the description a little "pushy": instead of "How to do X", write "How to do X. Use this skill whenever the user mentions X, Y, or Z, even if they don't explicitly ask for it."
• the rest of the skill body

Skill Writing Guide

Anatomy of a Skill

skill-name/
├── SKILL.md (required)
│   ├── YAML frontmatter (name, description required)
│   └── Markdown instructions
└── Bundled Resources (optional)
    ├── scripts/    - Executable code for deterministic/repetitive tasks
    ├── references/ - Docs loaded into context as needed
    └── assets/     - Files used in output (templates, icons, fonts)

Progressive Disclosure

Skills use a three-level loading system:

1. Metadata (name + description) — Always in context (~100 words)
2. SKILL.md body — In context whenever skill triggers (300 lines), include a table of contents

Domain organization — when a skill supports multiple variants, organize by variant:

cloud-deploy/
├── SKILL.md (workflow + selection logic)
└── references/
    ├── aws.md
    ├── gcp.md
    └── azure.md

Writing Patterns

Use the imperative form. Explain the why behind instructions — models understand reasoning, not just rules. Avoid heavy-handed ALL-CAPS MUSTs; explain why something matters instead.

Defining output formats:

## Report structure
Use this exact template:
# [Title]
## Executive summary
## Key findings
## Recommendations

Examples pattern:

## Commit message format
Input: Added user authentication with JWT tokens
Output: feat(auth): implement JWT-based authentication

Principle of Lack of Surprise

Skills must not contain malware, exploit code, or content that could compromise system security. A skill's contents should not surprise the user given its description.

Writing Style

Write a draft, then review with fresh eyes. Generalize from specific examples — don't overfit to one case. Keep skills lean: remove instructions that aren't pulling their weight. Read as if you're the model using this skill for the first time.

Validation

After writing a skill, run the quick validator to check structure:

python environment/skills//scripts/quick_validate.py environment/skills//

---

Output path for this task

Save each skill as SKILL.md inside a named subdirectory under environment/skills/ (relative path from your working directory). Do not write to /root/.claude/skills/ or any other location.

environment/skills//SKILL.md

Each SKILL.md must begin with the YAML frontmatter block (name + description).

After creating the skills, solve the task using them as reference.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: cxcscmu
• Source: cxcscmu/SkillLearnBench
• License: MIT
• Homepage: https://cxcscmu.github.io/SkillLearnBench

Install and usage instructions live in the source repository linked above.