skill-review_skill

Overview

This skill reviews and validates agent skills against a compiled set of best practices and standards. It inspects structure, frontmatter metadata, description quality, progressive disclosure, and common anti-patterns. The goal is actionable feedback and a short remediation plan to make skills reliable and triggerable.

How this skill works

The skill loads the standard references and a checklist, then identifies the review target (single skill file, all skills in a directory, or a draft). It runs a structural audit, validates frontmatter metadata, evaluates description trigger quality and token efficiency, scans for anti-patterns, and produces a concise report with prioritized recommendations. A quick-review mode can run only the checklist and return failures.

When to use it

• When you ask “review this skill”, “check my skill”, or “validate skill”
• When creating a new skill draft and you want feedback before publishing
• When you suspect poor trigger reliability or vague descriptions
• When you need a fast checklist-based audit across many skills
• When you want assurance that scripts, references, and paths are consistent

Best practices

• Keep the primary skill document short and put long details in referenced files
• Make the description third-person, include what it does and explicit trigger phrases
• Use forward-slash paths and document external binaries or dependencies
• Provide defaults for multi-option tasks and reduce degrees of freedom for fragile operations
• Prefer examples over long prose and keep each line justifying its token cost

Example use cases

• Validate a newly drafted skill file before adding it to the catalog
• Audit an entire skills directory for naming, path, and reference depth issues
• Quick-check a skill when a model fails to trigger it reliably
• Get a prioritized remediation list for a skill flagged as verbose or ambiguous
• Confirm that script paths referenced in the skill exist and are portable