🧠 ai-quality

Testing LLM features, guardrails, groundedness and agent skills.

Install: ai-quality@jaktestowac. See the quick start for each tool.

📘 Guide: AI product QA shows how the pieces work together.

Skills

• testing-llm-features
• testing-llm-guardrails
• reviewing-ai-output-groundedness
• testing-agent-skills
• unslop-answers

Commands

Run these yourself: /name in Claude Code and GitHub Copilot, $name in Codex. The agent never starts them on its own.

• design-llm-eval-suite: Design an LLM eval suite
• test-llm-guardrails: Test LLM guardrails and injection resistance

Agents

Agent	Claude Code	GitHub Copilot	Codex
LLM Feature Test Engineer - evals, guardrails, red team	llm-feature-test-engineer	llm-feature-test-engineer	llm_feature_test_engineer

Claude Code and GitHub Copilot load these agents with the plugin. Codex does not load agents from plugins: copy the files from codex/agents/ into your project's .codex/agents/ or into ~/.codex/agents/.