Authoring Kit

Skill Authoring

This page is designed for both humans and agents. Use the links below to fetch the raw Markdown and apply it directly when generating new skills.

Ready to test your draft? Scan it on /submit or return to the API docs for automation.

Best Practices

What the scanner flags and how to avoid it.

https://agentverus.ai/resources/skill-authoring-best-practices.md

SKILL.md Template

Copy/paste templates (OpenClaw, Claude, and generic).

https://agentverus.ai/resources/skill-template.md

Skill Writer (Agent Skill)

A skill definition that helps agents author safer skills.

https://agentverus.ai/resources/skills/agentverus-skill-writer/SKILL.md

Fast Validation

Run the scanner locally before publishing:

npx agentverus-scanner scan ./SKILL.md --json

Goal: no critical findings, minimal high findings, and a badge of CERTIFIED or CONDITIONAL.

State the exact purpose and scope in the description.
Use least-privilege tools and permissions; avoid broad or destructive access.
Add a Safety Boundaries section with explicit “must not” bullets.
Define output format constraints (JSON/Markdown shape, length limits).
Document error handling paths (missing inputs, failed calls, retries with bounds).
Avoid hidden instructions, obfuscation, and instruction-override language.