Skill Development Master Standard

Problem

Agent Skill delivery often breaks on small details: too many YAML fields, reliance on environment variables, multiple entry points, unverified API paths, or SKILL.md missing from the zip root. These issues may pass a local test but fail later during loading, distribution, or replay. This skill turns the development process into a checkable standard, narrowing the gap between a one-time run and a shippable package.

How It Works

It guides development step by step: confirm the skill name, core capability, output format, and external dependencies; then generate a requirements note, directory layout, and SKILL.md. Key rules include:

• Keep the YAML header to name and description only
• Store config in resources/config.json
• Use one execution path: curl or a Python script
• Verify API endpoints before delivery
• Ensure the zip root contains SKILL.md

It also requires test cases for normal, edge, out-of-scope, and failure scenarios, plus a security review for keys, permissions, logging, and compliance.

Boundaries

Use this for standardizing Agent Skill development, not for vague one-off scripts. Split overly complex features into sub-skills, pause when API endpoints cannot be verified, and keep each skill to a single responsibility with reproducible input and output.