skill-creator_skill

Overview

This skill teaches how to design and build effective agent skills that extend an AI with domain workflows, tooling, and reusable assets. It provides a clear process from understanding concrete examples to packaging and iterating a finished skill. The guidance emphasizes concise context, progressive disclosure, and practical reuse.

How this skill works

The skill inspects common task patterns and maps them to reusable components: a small manifest (name + description) that tells the agent when to use the skill, a concise guidance body that loads only on trigger, and optional bundled resources (scripts, references, assets) that the agent pulls in as needed. It recommends a three-tier loading model to minimize context cost and prescribes content organization patterns so the agent reads only relevant details.

When to use it

• Creating a new skill to encode a repeatable workflow or company policy
• Updating an existing skill to add scripts, templates, or domain references
• Packaging deterministic operations into scripts to avoid token-heavy repetition
• Designing multi-variant skills that must remain context-efficient
• Preparing skills for use by other AI agents that will pick them by manifest triggers

Best practices

• Keep the manifest short and explicit about when to trigger the skill and what it does
• Prefer concise examples over long prose; only add context the AI cannot infer
• Use progressive disclosure: manifest -> guidance body -> bundled resources on demand
• Choose freedom level to match task fragility: high for open tasks, low for brittle sequences
• Store deterministic code as executable scripts rather than long inline snippets
• Organize references by topic and link them from the guidance body to avoid deep nesting

Example use cases

• PDF processing skill with a rotate/ocr script and a reference of common formats
• Company policy skill containing canonical templates and a quick trigger description
• Frontend scaffold skill bundling HTML/React templates in references for rapid project starts
• Analytics skill that carries table schemas and query examples so agents can answer data questions reliably
• Document editing skill that describes triggers for creation, revision, tracked changes, and when to load deep format references