skill-authoring

Write portable agent skills following the agentskills.io open standard.

Updated May 29, 2026
One-click install
npx skills add https://github.com/RatnaAtika/Indonesian-ASR-11-Class-Dataset --skill skill-authoring-ratnaatika
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skill-authoring
Source: https://github.com/RatnaAtika/Indonesian-ASR-11-Class-Dataset/tree/main/.agents/skills/skill-authoring
Command: npx skills add https://github.com/RatnaAtika/Indonesian-ASR-11-Class-Dataset --skill skill-authoring-ratnaatika

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Agent skills often fail to load or behave inconsistently across harnesses because of malformed frontmatter, vague descriptions, or harness-specific content baked into the skill body. This Skill provides a complete authoring procedure so new skills are discoverable, portable, and versioned correctly. ## Core Features & Use Cases - Frontmatter contract: Enforces required fields (name, description, provides, version) with hard rules like name matching the directory and SemVer versioning. - Content structure template: Provides a proven skeleton (When to use, Workflow, Hard rules, Adaptation rules, Verification) with direct-voice style guidance. - Cross-harness portability: Explains how to keep SKILL.md canonical and delegate Cursor, Continue, Aider, and other adapters to a separate compatibility skill. - Use Case: A contributor opens a PR adding a new skill to a shared pack; use this Skill to review frontmatter validity, workflow testability, and the verification checklist before sign-off. ## Quick Start Use the skill-authoring skill to create a new portable agent skill for my repository with valid frontmatter and a verification checklist.

Frequently Asked Questions about skill-authoring

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write a SKILL.md file for an agent skill?

Create a SKILL.md with YAML frontmatter containing name, description, provides, and version, followed by a Markdown body with When to use, Workflow, Hard rules, and Verification sections. The name must match the directory name and the description must state when the agent should load it.

What frontmatter fields are required in an agent skill?

The required fields are name (lowercase-hyphenated, matching the directory), description (one sentence on when to use it), provides (a stable capability tag), and version (SemVer). Optional fields include license, categories, maintainers, and keywords.

Does one SKILL.md work across Claude Code, Cursor, and Aider?

The canonical SKILL.md works for Codex, Pi, Claude Code, and OpenCode directly. For Cursor, Continue, Cline, and Roo you write .mdc or .md adapters via a harness-compatibility skill, and Aider uses a concat adapter. Harness-specific instructions must stay out of the SKILL.md body.

How should I version an agent skill when updating it?

Use SemVer: patch bump for typo or link fixes, minor bump for new sections or capabilities that do not break callers, and major bump when a hard rule changes or the layout breaks downstream consumers. Keep the version monotonically increasing so smart-sync upgrade paths work.

What are common mistakes when authoring agent skills?

Common mistakes include descriptions that do not answer when to load the skill, hardcoded paths instead of cross-skill name references, harness-specific content embedded in the body, and nested skill layouts that most harnesses do not support.