Skill UI

Build automated evaluation suites for AI agents using golden datasets, rubrics, and regression gates. Covers prompt-level unit evals, tool-calling checks, end-to-end tasks, safety/adversarial tests, and LLM-as-judge scoring to gate deployments on quality.

Audit and optimize websites for platform-specific AI search engines including Google AI Overviews, ChatGPT, Perplexity, Gemini, and Bing Copilot. Provides per-platform checklists, scoring rubrics, and action items to improve visibility and citation rates in generative engine optimization (GEO). Requires network access for site audits; PDF reports need pandoc and headless Chrome.

This skill evaluates supplied outputs against explicit criteria, including code-change reviews and authorized safety evaluations. It supports evidence-backed review leads, rubric judgments, and batch transcript evaluation for existing outputs or observed behavior. It helps users judge results without merging or running targets. Suitable for design requests, code reviews, and safety workflows. Users can choose real API calls or simulation modes for independent outcome validation.

This skill evaluates code diffs against a specific rubric to determine issue detection and resolution. It reads diff files, rubrics, and original fixtures, then answers detection and fix questions for specific problems. Results are appended to a CSV file. Ideal for automated testing and quality assurance workflows in software development.