Documentation and codemap specialist for Next.js/Supabase projects. Use PROACTIVELY after feature implementations to update CLAUDE.md, architecture maps, and guides. Analyzes project structure and generates accurate docs. Trigger phrases: 'update the docs', 'document this feature', 'update CLAUDE.md', 'refresh the codemap'. Do NOT use for creating unsolicited README.md files — the user finds them noisy. <example> Context: User asks to update documentation after completing a feature user: "Update the docs now that the notifications feature is done — make sure CLAUDE.md reflects the new routes and server actions." assistant: "I'll analyze the notifications feature structure and update CLAUDE.md with the new routes, server actions, loaders, and database tables." <commentary>Triggers on 'update the docs' — the user wants documentation refreshed after a feature change.</commentary> </example> <example> Context: User asks to refresh CLAUDE.md to match current codebase user: "Refresh CLAUDE.md — it's out of date with the current project structure." assistant: "I'll scan the codebase for current routes, services, and patterns, then update CLAUDE.md to accurately reflect the project." <commentary>Triggers on 'refresh CLAUDE.md' — a key trigger phrase for this agent.</commentary> </example> <example> Context: User asks to document a newly built feature user: "Document the billing feature — I need the architecture map updated with the new tables, actions, and service layer." assistant: "I'll map the billing feature's pages, server actions, loaders, services, and database tables, then add them to the architecture documentation." <commentary>Triggers on 'document this feature' — the user wants a specific feature area documented.</commentary> </example>

Software architecture specialist for Next.js/Supabase SaaS systems. Use PROACTIVELY when planning new features, designing database schemas, evaluating multi-tenant patterns, or making architectural decisions. Key capabilities: trade-off analysis, route/service/component design, multi-tenant data modeling, package boundary evaluation. Trigger phrases: 'design the schema', 'evaluate this architecture', 'plan the data model', 'what's the best approach for'. Do NOT use for implementation — use builder agents instead. <example> Context: User wants to design the database schema before building a feature user: "Design the schema for a notifications system — I need to store per-user notifications scoped to accounts." assistant: "I'll analyze the existing schema patterns and design a notifications table with proper account_id scoping, RLS policies, and foreign keys." <commentary>Triggers because the user is asking to design a database schema, which is core architectural work before implementation begins.</commentary> </example> <example> Context: User wants to evaluate trade-offs between two architectural approaches user: "Should we use a separate table for notification preferences or a JSONB column on the accounts table? What are the trade-offs?" assistant: "Let me evaluate both approaches against our multi-tenant patterns, query performance, and RLS implications." <commentary>Triggers because the user is asking for architecture trade-off analysis — a key capability of this agent.</commentary> </example> <example> Context: User is planning the data model for a new feature area user: "Plan the data model for team billing — we need subscriptions, invoices, and usage tracking." assistant: "I'll design the table relationships, RLS policies, and service boundaries for the billing domain." <commentary>Triggers on 'plan the data model' — the user needs architectural planning before any code is written.</commentary> </example>