Staging repo for claude-code skills, commands, and agents. Deployable content lives in claude-files/.
- When working on skills (anything in folder
./claude-files/skills), always use the skill "Skill Development"
claude-files/ # Deployable to ~/.claude
├── CLAUDE.md # Project standards (for target codebases)
├── skills/ # /skill-name triggers
│ └── <skill-name>/
│ └── SKILL.md
└── commands/ # Slash commands
ccd/ # Deployment tool (Go)
deploy # Compiled binary
config.yaml # Deployment config
Each skill follows this structure:
skill-name/
├── SKILL.md # Required - YAML frontmatter + instructions
├── assets/ # Optional - Files used in OUTPUT (templates, images, boilerplate)
│ └── TEMPLATE.md # Example: document templates that get filled/copied
├── references/ # Optional - Documentation loaded INTO CONTEXT as needed
│ ├── patterns.md # Example: detailed patterns documentation
│ └── advanced.md # Example: advanced techniques
├── examples/ # Optional - Working code examples
│ └── example.sh
└── scripts/ # Optional - Executable utilities
└── validate.sh
| Directory | Purpose | Loaded Into Context? |
|---|---|---|
assets/ |
Files used in output (templates, images, boilerplate) | No - copied/used in output |
references/ |
Documentation Claude reads when needed | Yes - on demand |
examples/ |
Working code examples users can copy | Yes - on demand |
scripts/ |
Executable utilities for automation | No - executed directly |
---
name: skill-name
version: 1.0.0
description: Third-person description with trigger phrases. This skill should be used when the user asks to "trigger phrase 1", "trigger phrase 2", or needs guidance on [specific topic].
---
# Skill Title
[Lean content - 1,500-2,000 words ideal, <3,000 max]
[Reference assets/references/examples/scripts as needed]Use your skill-development skill to author skills in compliance for claude-code.
We have a tool writen in golang living in ccd folder that we use to manage the deployments of the claude-files to the target folders (eg. ~/.claude)