claude-skills

A collection of curated Claude Skills — reusable instruction packages that teach Claude how to perform specific tasks in a repeatable, grounded way.

Skills work across Claude.ai, Claude Code, and the Claude API. Build once, use everywhere.

Skills

Skill	Description	Version	Install
spec-forge	Ingest any API spec or SDK docs and become a grounded expert — generate code, validate integrations, diff versions	1.0.0	📦 Download
rosetta	Automate i18n workflows — extract strings, audit coverage, validate consistency, set up localization from scratch	1.0.0	📦 Download
zip-output	Ensures ZIP files always unzip into a clean single root folder — never loose files at archive root	1.0.0	📦 Download

Getting Started

Install a skill in Claude.ai

1. Download the .skill file from the table above
2. Go to Settings → Skills → Install from file
3. Upload the .skill file
4. The skill activates automatically when relevant

Install a skill in Claude Code

# Copy a single skill into your project
cp -r skills/spec-forge .claude/skills/spec-forge

# Or install all skills at once
npm run install-all

This copies every skill folder into .claude/skills/ in your current working directory. Restart Claude Code or run /reload-plugins to activate.

Repo Structure

claude-skills/
├── skills/                        # One folder per skill
│   ├── spec-forge/
│   │   ├── SKILL.md               # Skill instructions + frontmatter (required)
│   │   ├── README.md              # Skill-level documentation (required)
│   │   ├── evals/
│   │   │   └── evals.json         # Test cases
│   │   └── references/            # Reference docs loaded on demand
│   │       ├── openapi-cheatsheet.md
│   │       ├── auth-patterns.md
│   │       └── code-gen-patterns.md
│   ├── rosetta/
│   │   ├── SKILL.md
│   │   ├── README.md
│   │   ├── evals/
│   │   │   └── evals.json
│   │   └── references/
│   │       ├── frameworks.md
│   │       ├── locale-formats.md
│   │       └── key-conventions.md
│   └── zip-output/
│       ├── SKILL.md
│       ├── README.md
│       ├── evals/
│       │   └── evals.json
│       ├── references/
│       │   └── edge-cases.md
│       └── scripts/
│           ├── build_zip.py
│           └── verify_zip.py
├── scripts/
│   ├── validate.js                # Validate all skill structures
│   ├── package.js                 # Package skills into .skill files
│   └── install-all.sh             # Install all skills into Claude Code
├── .github/
│   └── workflows/
│       └── validate.yml           # CI: validate + package on push
├── README.md
├── CONTRIBUTING.md
├── LICENSE
└── package.json

Building a New Skill

1. Create a new folder under skills/
2. Add a SKILL.md with YAML frontmatter:

---
name: your-skill-name
metadata:
  author: "@shiftEscape"
  version: "1.0.0"
  tags: [tag1, tag2]
description: >
  What this skill does and when to trigger it. Be specific — this
  description is how Claude decides whether to load your skill.
---

# Your Skill Name

[Instructions here...]

3. Add a README.md describing the skill for the repo
4. Add test cases to evals/evals.json
5. Add reference files to references/ if needed
6. Run npm run package -- your-skill-name to generate the .skill file
7. Update the Skills table in this README
8. Open a PR — CI will validate automatically

Scripts

Command	Description
npm run package -- <skill>	Package a single skill into a .skill file
npm run package:all	Package all skills
npm run validate	Validate all skill structures
npm run install-all	Install all skills into .claude/skills/

Validation

The validator checks every skill for:

Hard errors (fail CI):

• Missing SKILL.md
• Missing or malformed YAML frontmatter
• Missing name or description field
• Invalid evals/evals.json (malformed JSON)

Warnings (logged, don't fail CI):

• Missing README.md
• Missing evals/evals.json
• Missing metadata.author
• Missing metadata.version

Run npm run validate -- --strict to treat warnings as errors.

License

MIT — see LICENSE