skill-evaluator_skill

Overview

This skill evaluates agent skills against Anthropic’s best practices for authoring and publishing. I produce structured evaluation reports with scores, clear findings, and actionable improvement recommendations. The output is designed to be used by authors and reviewers to prepare skills for publication and consumption by AI agents.

How this skill works

I run automated checks against naming conventions, description quality, file references, and manifest structure, then apply a manual rubric across dimensions like content quality, organization, and degrees of freedom. I identify anti-patterns and deduct points for common faults, then generate a concise report with weighted scores, strengths, and prioritized fixes. Reports include a checklist to guide pre-publication validation and testing.

When to use it

• Before publishing a skill to ensure it follows recommended conventions and is agent-ready
• When performing a peer review or audit of a skill’s content and structure
• To validate naming, descriptions, and file-reference hygiene before release
• When you need an objective, repeatable evaluation for quality gates
• If you want prioritized, actionable remediation steps rather than vague feedback

Best practices

• Use descriptive, gerund-based names (lowercase, hyphens) that reveal purpose and trigger patterns
• Write descriptions that state what the skill does and when to use it, in third person
• Keep documentation concise; remove explanations an intelligent agent already understands
• Organize content with progressive disclosure and limit long files; include short tables of contents when needed
• Calibrate degrees of freedom to task type: high for open text tasks, low for fragile scripted operations
• Avoid time-sensitive statements, inconsistent terminology, Windows-style paths, and deep nested references

Example use cases

• Automated pre-flight check for skills before a CI publish step
• Manual audit during a skills marketplace submission review
• Quality gate in a pull request to enforce naming and description rules
• Training authors on common anti-patterns with concrete remediation examples
• Batch evaluation of a skills collection to prioritize updates and fixes