Plugin 9.2: best practices for building skills
Plugin 9.2 adds one skill, applying-skill-best-practices. It holds the standards for building a good Agent Skill, gathered from Anthropic, OpenAI, AgentSkills.io, and Google and checked against each other, so you don’t have to read four sets of guidance before you build.
What it covers
Section titled “What it covers”Six decisions, grouped by when they come up:
- Before you build: check whether a skill for the job already exists, and scope the new one to a single job, built from two or three real examples of the task.
- While you build: write a description that triggers, keep the instructions to what the AI wouldn’t know, and choose the settings (whether it runs automatically, what tools it can use, and, on Claude Code, its model and effort).
- Before you trust it: prepare test cases from your real examples, run them with and without the skill, and check that it fires when it should.
How to use it
Section titled “How to use it”It doesn’t build skills itself. Name it alongside your platform’s skill creator:
Use the skill creator and follow applying-skill-best-practices to build a skill for [task]. Here are two real examples of the task: …
On Claude Code, Anthropic’s skill-creator runs the tests for you. On other platforms, hand the skill creator your examples first; where it can’t run the tests, the skill’s grading prompt and test-report format let you run them by hand.
Install or update the plugin, or download applying-skill-best-practices.zip.