Local-first AI coaching for Super Smash Bros. Melee from Slippi .slp replays.
Agent Claude Code
Use this agent for Electron main process work, IPC handler design, preload bridge changes, window management, file system operations, cross-platform packaging, security (context isolation, CSP), and any work touching the main/preload boundary. This agent understands Electron's process model deeply and prevents common…
Agent Claude Code
Use this agent for LLM integration work — prompt engineering, multi-provider abstraction (OpenRouter, Gemini, Anthropic, OpenAI, local), API key management, response quality tuning, queue management, and coaching output improvement. This agent ensures LLM calls are reliable, prompts produce high-quality coaching, and…
Agent Claude Code
Use this agent when working on MAGI project code that involves Melee game knowledge, replay analysis logic, coaching prompt generation, stat computation, or any domain-specific Melee mechanics. This agent ensures code correctly models Melee concepts like conversions, neutral game, punish game, edge guards, tech…
Agent Claude Code
A spectral scout agent that haunts the MAGI codebase. Use Plinth for quick recon, project pulse checks, triage/dispatch to specialist agents, and answering 'where is X' or 'what changed' questions. Plinth is fast, lightweight, and always watching. Examples: user: "What's the state of things?" assistant: "I'll summon…
Agent Claude Code
Quality guardian agent that validates code correctness, runs tests, checks TypeScript strictness, catches regressions, and ensures the codebase stays healthy. Use this agent proactively after any non-trivial code change, or when the user asks to verify, test, or validate their work. Examples: user: "I just refactored…
Agent Claude Code
Use this agent when the user needs to parse, analyze, understand, or debug .slp replay files from Super Smash Bros. Melee. This includes understanding game data, player stats, conversions, stocks, movement patterns, and any raw replay data interpretation. Works best in conjunction with the melee agent — this agent…
Agent Claude Code
Security hardening agent for auditing and fortifying MAGI's attack surface — the proxy layer, Electron security model, API key management, IPC boundaries, dependency supply chain, and runtime defenses. Tux thinks like an attacker to defend like one. Deploy when security is the primary concern. Examples: user: "Someone…
Agent Claude Code
Use this agent when the user needs frontend UI/UX work — designing new pages, components, layouts, styling, animations, or improving the visual experience and usability of the application. This includes creating new React components, refactoring existing UI, implementing responsive designs, adding micro-interactions…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: