nikita

Best-practice prompt engineering using research-backed patterns. Handles: transforming rough prompts into production-ready, debugging prompt issues, creating new prompts from scratch. ALWAYS researches via subagents (keeps main context clean). Triggers on: "improve prompt", "fix prompt", "create prompt", "prompt not working", "write a prompt for", "optimize prompt", "/prompt".

Rigorous TDD workflow enforcement with Red-Green-Refactor cycle, deployment verification, and log monitoring. Use when fixing bugs, implementing features, or verifying deployments.

Complete user journey simulation for the Nikita AI girlfriend game. Simulates a real player (Simon) playing through all 5 chapters — registration, onboarding, text gameplay, boss encounters, decay, engagement states, vice personalization, voice calls, portal monitoring, terminal states, and adversarial scenarios. Assesses behavioral realism (does Nikita feel human?), game balance (are thresholds achievable?), and classifies all findings by severity and category. Use this skill whenever the user says "e2e", "end to end test", "full journey test", "test nikita", "regression test", "verify deployment", "test the bot", "test onboarding", "simulate a game", "run the test suite", "play through the game", "test the experience", "check if nikita feels real", "behavioral test", "game balance check", or after any deployment to production. This is THE authoritative E2E testing method — replaces the archived e2e-test-automation and e2e-journey skills. Always use /e2e, never the old commands.

Execute end-to-end tests for Nikita using Telegram MCP, Gmail MCP, Supabase MCP, Chrome DevTools MCP, and gcloud CLI. Use when verifying implementations, testing user journeys, validating integrations, performing regression testing, or after completing any feature implementation. MANDATORY after /implement completes.