Find Skills for Agents

Problem This Skill Addresses

When a user asks whether there is a skill for a task or how to complete a recurring workflow, an agent should not rely only on its currently loaded capabilities. It should also check whether the open agent-skills ecosystem already contains a more suitable extension. test-find-skills turns natural-language requests into searchable skill keywords. For example, a request about “React app performance optimization” can be reduced to react and performance, then used to shortlist reusable workflows, tools, or templates. This helps avoid turning a one-off question into a custom long-term task script and reduces the need to manually compare multiple skill repositories.

How It Works

The core job of this skill is skill discovery and recommendation, not direct execution of a specific domain task. It usually works in three steps:

• Identify the domain, the concrete task, and whether the task is common enough that an existing skill is likely to exist.
• Search the open skill ecosystem for candidates across categories such as web development, testing, DevOps, documentation, code quality, design, and workflow automation.
• Present each relevant option with a plain explanation of what it does, how to obtain it, and where to inspect its details.

It favors more specific search terms, such as nextjs, playwright, ci-cd, and api-docs, rather than overly generic keywords. If no suitable skill is found, it should say so explicitly, then fall back to helping with the task directly or suggest creating a custom skill based on the user’s needs.

Boundaries and Caveats

This skill is best for meta-tasks such as finding tools, workflows, or templates, not for replacing the execution layer of a specialized skill. Its recommendation quality depends on how clearly the user defines the problem. A vague request like “help me optimize my website” may need follow-up about the tech stack, performance bottleneck, and deployment environment before it becomes a useful search. Open skills may also come from different repositories, so their permission scope, dependencies, and maintainability still need review. For workflows involving credentials, production environments, or internal data, the existence of a similar skill in the ecosystem is not enough to justify enabling it without further validation.