Skills Development

This repo contains Claude Code skills - agent-invoked tools for specialized tasks.

Follows the Agent Skills Specification.

Structure

skills/
├── CLAUDE.md                    # This file
└── <skill-name>/
    ├── SKILL.md                 # Required manifest
    ├── .env.example             # Env var template
    └── scripts/
        └── <script>             # uv script with CLI (no .py extension)

Creating a Skill

1. Directory

mkdir -p skills/<skill-name>/scripts

2. Script (scripts/<script>)

uv script with shebang and inline deps:

#!/usr/bin/env -S uv run --script
# /// script
# requires-python = ">=3.11"
# dependencies = [
#     "some-client",
#     "python-dotenv",
#     "typer",
# ]
# ///

Requirements:

• Use typer for CLI
• Use python-dotenv for env loading
• Default output: JSON (for LLM consumption)
• Support --format table for human review
• Load credentials from env vars only (no .envrc parsing)
• Make executable: chmod +x

3. SKILL.md

---
name: skill-name
description: What it does. When to use it (trigger keywords).
allowed-tools: Bash, Read
---

# Skill Name

## When to Use
- Trigger conditions...

## Requirements
- `ENV_VAR` - description

## Command
\`\`\`bash
./scripts/script [command] [options]
\`\`\`

## Options
| Option | Description |
|--------|-------------|
| `-o` | ... |

## Output Format
JSON structure...

## Examples
...

4. .env.example

# Service API Key
SERVICE_API_KEY=your_key_here

Conventions

• Script filenames: snake_case (no .py extension)
• Skill directories: kebab-case
• Functions/methods: camelCase
• Default output: JSON for agents
• Concise excerpts for token efficiency
• Error handling with typer.Exit(1)

Testing Locally

# Symlink to personal skills
ln -s /path/to/skills/<skill-name> ~/.claude/skills/<skill-name>

# Restart Claude Code to reload
# Ask: "What Skills are available?"

Existing Skills

See the tables in README.md (single source of truth; update it when adding or removing a skill).