Contributing to Agent Powerups

Keep contributions small, explicit, and portable.

Rules

• Do not add fake compatibility claims.
• Do not add hidden installers.
• Do not add secrets, machine-specific paths, or personal data.
• Prefer one strong asset over several thin ones.

Skill Requirements

Each skill folder under skills/ must include SKILL.md with YAML frontmatter:

---
name: skill-name
description: Use when ...
---

Use YAML frontmatter plus a pure Markdown body for SKILL.md, agent instruction files, templates, examples, and writing guidelines. Prefer Markdown headings such as ## Purpose, ## When to Use, ## Workflow, and ## Verification. Do not use XML-like tags such as <Purpose> or <Workflow> as default top-level sectioning; reserve them only for strict delimiters around nested examples, quoted inputs, external documents, or machine-readable prompt payloads.

Tool-dependent skills must also document:

• required command or package
• how to check if it exists
• how to install it
• that install requires user approval
• fallback behavior if install is declined

Catalog Requirements

Every shipped asset needs a catalog.json entry.

Allowed type values:

• skill
• command
• mcp-config
• agents-md-template
• hook
• workflow
• example
• script
• pack

Allowed maturity values:

• draft
• beta
• stable

Validation

Run before opening a pull request:

python scripts\validate-skills.py
python scripts\validate-catalog.py
python scripts\check-requirements.py

Pull Request Checklist

• Skill files are non-empty and have name + description
• Tool-dependent skills document install/check/fallback behavior
• catalog.json was updated if shipped assets changed
• Validation scripts pass
• Compatibility claims are narrow and defensible