claude-skills_skill

Overview

This skill turns scattered domain material (docs, APIs, code, specs) into a reusable, maintainable Skill package with clear activation and quality checks. It scaffolds a decidable SKILL.md entrypoint, moves long-form evidence to references/, and adds scripts/assets and a delivery quality gate. Use it to create, refactor, or validate Skills so they reliably trigger and remain maintainable.

How this skill works

I inspect source material (docs, repos, code samples) to identify actionable patterns, triggers, and required inputs. I generate a concrete directory layout, an actionable SKILL.md with YAML frontmatter and Quick Reference patterns, and a references/ folder with a navigable index. I add a pre-delivery checklist and wiring for local validation scripts to enforce activation rules and quality gates.

When to use it

• Creating a new Skill from docs, specs, or repo material
• Refactoring an existing Skill that is too long, unclear, or misfires
• Designing reliable activation with frontmatter, triggers, and boundaries
• Extracting a concise Quick Reference from long-form content
• Adding a quality gate and validator before delivery

Best practices

• Write frontmatter description as "what + when" with concrete trigger keywords
• Keep Quick Reference patterns short and directly usable (<= 20 items)
• Move long explanations to references/ and provide references/index.md for navigation
• Define "Not For / Boundaries" and required inputs to prevent misfires
• Provide >= 3 reproducible end-to-end examples with acceptance criteria

Example use cases

• Generate a minimal Skill scaffold (skill-name/ with SKILL.md) from an official API doc and 2–3 code samples
• Refactor a doc-dump SKILL.md by extracting patterns into Quick Reference and moving long text to references/
• Add validation and a pre-delivery checklist to a Skill and run validate-skill.sh --strict before shipping
• Split a large knowledge base into references/ sections and create a concise SKILL.md that triggers reliably