# Capability Boundary — v1.1.0-semantic

> Frozen: 2026-06-05T19:40:00+08:00
> This document defines what the system CAN and CANNOT do. Do not claim capabilities outside this boundary.

## ✅ CAN DO (Certified)

### Web Fullstack Generation
- **Input:** Chinese natural language requirement (1 sentence)
- **Output:** Complete Next.js + Express monorepo
- **Quality:** Domain-specific tables, English identifiers, CRUD APIs, auth scaffolding
- **Verified:** 18 domains, 170 regression checks

### Electron Desktop Packaging
- **Input:** Fullstack monorepo
- **Output:** Electron 35 app with tray, IPC, auto-updater skeleton
- **Platforms:** Windows, macOS, Linux

### Release Packaging
- **Input:** Fullstack monorepo
- **Output:** Version manifest, SHA256 checksums, platform installers (placeholders)

### Semantic Generation
- **Chinese → English:** API paths, table names, route names, service names
- **Field specificity:** 60+ domain-specific field templates
- **Identifier safety:** Singularize exceptions, longest-match slug, exclude rules

## ⚠️ CANNOT GUARANTEE (Not Certified)

### Native Mobile
- **iOS (SwiftUI)** — Tech stack exists in template but no builder agent
- **Android (Jetpack Compose)** — Tech stack exists in template but no builder agent
- **React Native / Flutter** — Mentioned in tech stack, not generated

### Mini Programs
- **微信小程序** — Platform detection works, no builder agent
- **uni-app** — Tech stack template only

### Backend Alternatives
- **NestJS** — Architecture suggests NestJS, actual generation uses Express
- **Prisma ORM** — Architecture suggests Prisma, actual uses better-sqlite3
- **PostgreSQL** — Architecture suggests PostgreSQL, actual uses SQLite

### Distributed Systems
- **Microservices** — Single-service architecture only
- **Message queues** — No Kafka/RabbitMQ integration
- **Service mesh** — No Istio/Linkerd support

### AI Integration
- **AI Agent clusters** — No multi-agent orchestration
- **LLM-powered features** — No AI feature generation
- **Smart recommendations** — Placeholder only

### Production Readiness
- **Load testing** — No performance benchmarks
- **Security audit** — No penetration testing
- **CI/CD pipelines** — GitHub Actions skeleton only
- **Monitoring/observability** — No APM integration

## Boundary Rules

1. **Do not add new agents** — SF-01 through SF-07 only
2. **Do not add new platforms** — Web + Electron + Release only
3. **Do not add new builders** — Current 7 agents only
4. **All changes must pass regression gate** — `node scripts/regression-gate.mjs`
5. **Capability expansion requires explicit unfreeze** — Document in this file