agent-skill-test

What problem does it solve?

Testing agent skills in shared or unstandardized environments causes cross-test contamination, inconsistent results across different CLI agents, and difficulty reproducing failures. This Skill eliminates that friction by providing a standardized, isolated workspace setup for skill testing.

Core Features & Use Cases

• Portable Isolated Workspaces: Creates self-contained test directories that copy skills under test to a universal skills-under-test/ path, with optional runner-native mirrors for Codex, Claude Code, Qoder CLI, and OpenCode.
• Runner-Agnostic Command Generation: Builds correct, runner-specific execution commands for multiple popular CLI agents, handling shell quoting, output capture, and permission flags automatically.
• Generic Validation & Evidence Capture: Includes scripts to capture runner events, transcripts, and final messages, plus validate required files, regex patterns, and forbidden artifacts to ensure test outcomes meet acceptance criteria.
• Use Case: Use this Skill to smoke test new skills, run regression tests after skill updates, validate skill behavior across different agent runners, or prepare portable skill copies for distribution.

Quick Start

Use the agent-skill-test skill to create an isolated test workspace for your target agent skill, run it with your preferred CLI agent, and validate the produced artifacts against your expected requirements.