Reverse Engineering Assistant | 反向工程助手

System archeology framework: reverse engineer existing systems across three dimensions — Logic, Data, and Runtime.

系統考古框架：從三個維度反向工程既有系統——邏輯、資料、執行環境。

Three Dimensions | 三大維度

┌─────────────────────────────────────────────────────────┐
│              System Archeology Framework                   │
├──────────┬──────────────┬────────────────────────────────┤
│  Logic   │     Data     │          Runtime               │
│ (spec)   │    (data)    │         (runtime)              │
├──────────┼──────────────┼────────────────────────────────┤
│ APIs     │ DB Schemas   │ Logs & Error Patterns          │
│ Modules  │ ORMs/Models  │ Config & Environment           │
│ Flows    │ Migrations   │ Metrics & Performance          │
│ Tests    │ Seed Data    │ Infra & Deployment             │
└──────────┴──────────────┴────────────────────────────────┘

Subcommands | 子命令

Full Analysis Mode | 全面分析模式

When /reverse is invoked without a subcommand, execute all three dimensions sequentially:

1. Data → Scan schemas, ORMs, migrations
2. Runtime → Analyze logs, configs, deployment
3. Logic (spec) → Extract APIs, flows, tests → Generate SPEC

Output: Integrated System Archeology Report combining all three dimensions.

Dimension Details | 維度詳情

spec: Logic Dimension (既有)

1. Scan - Read source files and identify public APIs, data flows, and business logic
2. Classify - Tag each finding as [Confirmed], [Inferred], or [Unknown]
3. Structure - Organize into SDD spec format with Acceptance Criteria
4. Attribute - Cite every reversed item with file:line source reference

data: Data Dimension (新增)

1. Discover - Find database schemas, ORM models, migration files, seed data
2. Map - Build entity-relationship model from code evidence
3. Classify - Tag relationships as [Confirmed] (FK constraints) or [Inferred] (code patterns)
4. Report - Output data model spec with:
   • Entity list with fields and types
   • Relationship map (1:1, 1:N, M:N)
   • Index and constraint inventory
   • Migration history summary
   • Data flow paths (write → read)

Evidence sources: schema.prisma, *.migration.*, models/, entities/, knexfile.*, sequelize, typeorm, SQL files, docker-compose.yml (DB services)

runtime: Runtime Dimension (新增)

1. Scan configs - Environment variables, config files, feature flags
2. Analyze logs - Log patterns, error frequency, log levels
3. Check infra - Docker configs, CI/CD pipelines, deployment manifests
4. Baseline - Output runtime baseline with:
   • Environment variable inventory (names only, never values/secrets)
   • Config file map and hierarchy
   • External service dependencies (APIs, queues, caches)
   • Deployment topology (containers, services)
   • Health check and monitoring endpoints

Evidence sources: .env.example, docker-compose.yml, Dockerfile, *.config.*, CI/CD files, k8s/, log files (patterns only)

Security: NEVER output actual secret values. Only list variable names and describe their purpose.

Anti-Hallucination Rules | 防幻覺規則

Usage | 使用方式

/reverse                              - Full 3-dimension analysis | 三維度全面分析
/reverse spec src/auth/               - Logic: extract spec | 邏輯：提取規格
/reverse data                         - Data: analyze schemas & models | 資料：分析結構
/reverse runtime                      - Runtime: analyze configs & infra | 執行環境：分析配置
/reverse bdd specs/SPEC-AUTH.md       - Convert spec ACs to Gherkin
/reverse tdd features/auth.feature    - Analyze test coverage

Next Steps Guidance | 下一步引導

After /reverse (full or spec) completes, the AI assistant should suggest:

系統考古完成。建議下一步 / System archeology complete. Suggested next steps:

• 執行 /sdd 審查並核准此規格 ⭐ Recommended / 推薦 — Review and approve the generated spec
• 執行 /derive 從規格推導測試 — Derive tests from spec (requires approval first)
• 審查規格中的 [Inferred] 和 [Unknown] 標記 — Review uncertainty tags manually

Reference | 參考

• BDD Extraction Workflow Guide — Read it when extracting BDD scenarios out of an existing spec.
• TDD Analysis Workflow Guide — Read it when analysing test coverage against scenarios and finding the gaps.
• Step-by-step workflow: workflow.md — the reverse-engineering phases in order (code scanning, data models, configuration, test analysis, coverage). Read it when actually running the process rather than deciding whether to.
• Detailed guide: guide.md
• Core standard: reverse-engineering-standards.md

AI Agent Behavior | AI 代理行為

完整的 AI 行為定義請參閱對應的命令文件：/reverse

For complete AI agent behavior definition, see the corresponding command file: /reverse