agent-skill-infra

Guide for writing behavior test cases (evals.json) for Agent Skills. Defines the eval schema, judgment criteria (trigger/flow/output), and snapshot testing conventions. Use when creating test cases for Skills or teaching others how to test Agent Skills.

Analyze and score Agent Skills using an 8-dimension quality framework (helloandy methodology). Checks trigger accuracy, output structure, error tolerance, token efficiency, and security posture. Use when reviewing, improving, or evaluating SKILL.md files for quality.

Track and manage Agent Skill versions using Git-backed versioning. Provides change summaries, behavior regression detection, dependency validation, and one-click rollback. Use when updating Skills, reviewing changes, or diagnosing post-update issues.