Ismael
Marin

High-fidelity evaluation engine for benchmarking AI agent skills across any stack (Rails-first, but extensible). Compare baseline vs. skill-enhanced agent runs with 100% reproducibility via isolated Git sandboxes.

• Side-by-Side Evaluation: Quantify the impact of specific context or skills on agent performance
• Isolated Git Sandboxes: Clean diffs, zero side-effects, and 100% reproducibility for every run
• Multi-Provider: Native support for Anthropic, Gemini, OpenAI, DeepSeek, Groq, Ollama, and more
• LLM-based blind judging across Correctness, Skill Adherence, Code Quality, and Test Coverage

Turn any Rails app into an AI-ready codebase — one command, zero config. AI assistants waste tokens guessing your app structure and still give generic responses. rails-ai-bridge fixes the root cause.

• Generates static context files for every AI tool — CLAUDE.md, AGENTS.md, .cursorrules, copilot-instructions.md, .windsurfrules, .mcp.json
• No server required — commit the files, whole team benefits automatically
• Live MCP server mode for real-time Rails introspection on demand
• Works with Claude Code, Cursor, Codex, GitHub Copilot & Windsurf

Curated AI agent skill library for Ruby on Rails development. Structured SKILL.md files that give coding agents the context, conventions, and discipline to generate production-quality Rails code.

• Tests as a hard gate — agents cannot write implementation before tests exist and fail
• Workflow chaining from PRD → tasks → test gate → implementation → review → PR
• 30+ skills covering service objects, engines, migrations, security & background jobs
• Compatible with Cursor, Codex, and Claude Code