Starchild Official Skills

Official skills maintained by the Starchild team. Installed via npx skills.

Building a new skill? See CONTRIBUTING.md for the full guide — directory structure, SKILL.md frontmatter, exports.py patterns, function design best practices, logo specs, and a publishing checklist.

Install a Skill

npx skills add Starchild-ai-agent/official-skills --skill hyperliquid

Or within a Starchild agent conversation:

search_skills(query="hyperliquid")    # searches + auto-installs

Repository Structure

official-skills/
├── skills.json              ← auto-generated index (do not edit manually)
├── .github/workflows/
│   └── build-index.yml      ← CI: validates + rebuilds skills.json on push
├── hyperliquid/
│   └── SKILL.md
├── birdeye/
│   ├── SKILL.md
│   ├── token.py             ← tool scripts (optional)
│   └── templates/
└── ...

Each skill lives in its own top-level directory. The only required file is SKILL.md.

SKILL.md Format

Every SKILL.md must have YAML frontmatter with three required fields:

---
name: my-skill
version: 1.0.0
description: Short summary of what this skill does
---

# My Skill

Instructions, usage examples, API references, etc.

Required Fields

Field	Rules	Example
name	Lowercase, alphanumeric + hyphens	hyperliquid
version	Semver	1.0.0
description	One-line summary for search	Trade perpetual futures on Hyperliquid DEX

Optional Fields

Field	Purpose	Example
tools	List of tool names this skill provides	[hl_account, hl_order]
metadata	Starchild-specific metadata (emoji, skillKey)	starchild: { emoji: "📊" }
user-invocable	Whether users can invoke this skill directly	true
tags	Search tags	[trading, defi]

Multi-File Skills

Skills can include additional files (Python scripts, templates, configs). Place them alongside SKILL.md in the same directory:

birdeye/
├── SKILL.md
├── __init__.py
├── token.py
├── wallet.py
└── tools/

npx skills add copies the entire directory recursively.

Development Workflow

The quick steps below cover the basics. For the full guide — including exports.py patterns, function design principles (one-function-per-intent, read+execute pairing), logo sourcing, and a publishing checklist — see CONTRIBUTING.md.

Add a New Skill

1. Create a directory:

mkdir my-new-skill

2. Write SKILL.md with the required frontmatter (name, version, description).

3. Push to main:

git add my-new-skill/
git commit -m "feat: add my-new-skill"
git push

4. GitHub Actions automatically validates the frontmatter and updates skills.json.

Update an Existing Skill

1. Edit the skill files.

2. Bump the version in the frontmatter:

version: 1.0.0  →  version: 1.1.0

3. Push. CI updates skills.json automatically.

Users who already installed the skill will get the new version on next npx skills add (npx detects the changed computedHash).

Remove a Skill

1. Delete the directory:

rm -rf old-skill
git add -A && git commit -m "chore: remove old-skill" && git push

2. CI updates skills.json automatically (skill removed from index).

3. To also remove from running user containers, add the skill name to config/skill-removals.txt in the starchild-clawd repo.

CI Pipeline

On every push to main that touches */SKILL.md:

1. Validate — scans all SKILL.md files, fails if any are missing name, version, or description
2. Rebuild — generates skills.json from frontmatter
3. Commit — if skills.json changed, auto-commits and pushes

skills.json is auto-generated. Do not edit it manually — your changes will be overwritten.

Conventions

• Naming: lowercase, hyphens only (e.g. trading-strategy, not TradingStrategy)
• Versioning: semver — bump patch for fixes, minor for new features, major for breaking changes
• One skill per directory: each directory = one installable unit
• Keep skills focused: one skill = one integration or workflow (e.g. hyperliquid, not all-exchanges)
• Description matters: it's the primary search field — make it specific and keyword-rich