ephemeral-wasm-sandbox-executor — vibes-plug Skill

English | Bahasa Indonesia

---

<a name="english"></a>

English

Orchestration & Integration

Connects and orchestrates with brainstorming, zero-to-prod-orchestrator, autonomous-tdd-debugger, wasm-edge-computing-expert, autonomous-red-teamer, and production-ready-hardener to safely execute untrusted AI-generated code, simulate database migrations in-memory, and run security fuzzers with zero host blast radius.

Description

Production guide for architecting ultra-low latency, zero-trust ephemeral sandboxes powered by WebAssembly (WASM) runtimes (such as Wasmtime, Extism, and PGlite/SQLite WASM). Provides agents with instant execution environments (<5ms startup overhead) to dry-run database migrations, test adversarial inputs, execute generated logic, and evaluate potential security hazards without Docker container overhead or risk to host systems.

Trigger Conditions

Activate this skill when:

• Testing AI-generated code snippets or regex patterns that could trigger infinite loops or ReDoS.
• Dry-running SQL migrations, schema alterations, or database seeds in an in-memory WASM Postgres/SQLite instance before staging.
• Executing dynamic user-submitted code in multi-tenant SaaS environments with strict CPU and memory quotas.
• Running high-speed fuzz tests and dynamic property-based tests across generated functions.

---

Core Concepts & Patterns

1. Sandbox Comparison Matrix

Runtime Model	Cold Start Latency	Memory Overhead	Host Isolation Level	Best Use Case
Traditional Docker / OCI	1,500ms - 4,000ms	100MB - 500MB	Hypervisor / Namespace	Heavy integration tests, full services
Node.js vm module	< 1ms	< 2MB	Weak (Prototype pollution vulnerable)	Not suitable for untrusted AI execution
WASM Micro-Runtime (Extism / Wasmtime)	< 3ms	< 10MB	Absolute (Memory-isolated linear memory)	AI code validation, logic verification, transforms
In-Memory WASM DB (PGlite)	< 15ms	< 30MB	Complete file-system isolation	Zero-risk SQL migration dry runs

2. In-Memory Migration Dry-Runner (PGlite Implementation)

import { PGlite } from '@electric-sql/pglite';

export interface MigrationDryRunResult {
  success: boolean;
  appliedSteps: number;
  executionTimeMs: number;
  error?: string;
}

export class EphemeralDatabaseSandbox {
  private db: PGlite | null = null;

  public async initialize(): Promise<void> {
    this.db = new PGlite();
    await this.db.waitReady;
  }

  public async dryRunMigrations(sqlStatements: string[]): Promise<MigrationDryRunResult> {
    if (!this.db) {
      await this.initialize();
    }

    const startTime = performance.now();
    let applied = 0;

    try {
      await this.db!.query('BEGIN;');

      for (const statement of sqlStatements) {
        if (!statement.trim()) continue;
        await this.db!.query(statement);
        applied++;
      }

      await this.db!.query('ROLLBACK;');

      return {
        success: true,
        appliedSteps: applied,
        executionTimeMs: performance.now() - startTime
      };
    } catch (err: unknown) {
      if (this.db) {
        await this.db.query('ROLLBACK;').catch(() => {});
      }
      return {
        success: false,
        appliedSteps: applied,
        executionTimeMs: performance.now() - startTime,
        error: err instanceof Error ? err.message : String(err)
      };
    } finally {
      if (this.db) {
        await this.db.close();
        this.db = null;
      }
    }
  }
}

3. Resource-Metered Sandboxing Architecture

To prevent adversarial infinite loops and excessive resource consumption:

1. Instruction Fuel Metering: Deduct fuel units per WASM instruction opcode; abort when fuel drops to zero.
2. Linear Memory Caps: Constrain guest WASM memory pages to a hard limit (e.g. 64MB max).
3. No Ambient System Calls: Deny disk writes, outbound sockets, and environment variable access by default unless explicitly shimmed through capability-based security.

---

Best Practices

1. Dry-Run All Migrations in WASM First: Before applying DDL migrations to staging or production databases, run the exact SQL script in an ephemeral PGlite instance to catch syntax errors and foreign key violations.
2. Enforce Strict Fuel Metering: Attach instruction fuel limits to AI-generated loops to prevent denial-of-service or hangs.
3. Isolate Guest State: Instantiate a fresh WASM memory instance per test invocation to guarantee zero cross-test state pollution.
4. Capture Standard Streams: Pipe guest stdout and stderr into structured telemetry rather than unbuffered process output.

---

Common Pitfalls to Avoid

Anti-Pattern	Vulnerability	Engineering Remedy
Using eval() or Node.js vm for AI code	Prototype escape and host environment compromise	Execute exclusively inside WASM or V8 isolate sandbox
Applying migrations directly to staging DB	Schema corruption and long downtime on failed rollback	Dry-run in ephemeral WASM Postgres before applying
Allowing unbounded WASM memory growth	Host OOM (Out Of Memory) crash	Configure max memory pages (memory.grow ceiling)

---

Integration with Other Skills (MANDATORY)

• autonomous-tdd-debugger — Use sandbox as the execution harness for autonomous test verification.
• wasm-edge-computing-expert — Share WASM compilation toolchains, runtime profiles, and memory limits.
• autonomous-red-teamer — Safely execute exploit payloads and dynamic fuzzing within isolated WASM boundaries.
• production-ready-hardener — Verify all sandbox security headers, fuel limits, and capability policies.

Referenced By Orchestrators (MANDATORY)

• brainstorming — Added to "Testing & Keamanan" and "Backend & Runtime" matrix rows.
• zero-to-prod-orchestrator — Integrated in Phase 6 (Automated Testing, Error Resilience & Security Audit).

---

<a name="bahasa-indonesia"></a>

Bahasa Indonesia

Integrasi Orkestrasi

Terhubung dan mengorkestrasi dengan brainstorming, zero-to-prod-orchestrator, autonomous-tdd-debugger, wasm-edge-computing-expert, autonomous-red-teamer, dan production-ready-hardener untuk mengeksekusi kode hasil AI secara aman, mensimulasikan migrasi database in-memory, dan menjalankan pengujian keamanan tanpa risiko pada sistem host.

Deskripsi

Panduan produksi untuk merancang sandbox efemeral berlatensi ultra-rendah dan berprinsip zero-trust yang ditenagai oleh runtime WebAssembly (WASM) seperti Wasmtime, Extism, serta PGlite/SQLite WASM. Menyediakan lingkungan eksekusi instan (<5ms) bagi agen otonom untuk memvalidasi migrasi database, menguji input berbahaya, dan memverifikasi kode hasil AI tanpa overhead kontainer Docker.

Kondisi Pemicu

Aktifkan skill ini ketika:

• Menguji kode hasil AI atau pola regex yang berpotensi memicu loop tak berujung atau kerentanan ReDoS.
• Melakukan dry-run skrip migrasi SQL pada database in-memory (PGlite WASM) sebelum diterapkan ke database staging.
• Menjalankan kode pengguna pada arsitektur multi-tenant dengan kuota CPU dan memori yang ketat.
• Melakukan pengujian fuzzing kecepatan tinggi untuk memvalidasi ketahanan fungsi.

---

Konsep Inti & Pola Praktik

1. Validasi Migrasi In-Memory (PGlite)

Sebelum menerapkan migrasi database apa pun:

1. Inisialisasi Cepat: Buat instance Postgres WASM in-memory dalam hitungan milidetik.
2. Eksekusi Transaksional: Jalankan seluruh skrip DDL migrasi di dalam blok transaksi.
3. Verifikasi Integritas: Pastikan foreign key, tipe data, dan constraint terpasang tanpa konflik.
4. Rollback & Buang: Buang instance WASM tanpa meninggalkan jejak atau mengubah disk host.

2. Pembatasan Sumber Daya (Fuel Metering)

• Penghitungan Instruksi: Tetapkan batas instruksi komputasi (fuel). Jika kode mengalami infinite loop, runtime WASM akan otomatis menghentikan eksekusi saat bahan bakar habis.
• Batas Memori Linear: Kunci alokasi memori maksimal pada angka aman (misalnya 64MB) guna mencegah crash Out of Memory pada server.

---

Praktik Terbaik

1. Dry-Run Seluruh Migrasi di WASM: Validasi sintaks SQL dan dependensi relasional pada PGlite terlebih dahulu.
2. Gunakan Batasan Bahan Bakar (Fuel): Hindari eksekusi tanpa batas waktu untuk kode yang dihasilkan oleh model AI.
3. Isolasi Mutlak State: Gunakan instance memori baru untuk setiap pengujian agar tidak terjadi kebocoran state antar pengujian.
4. Blokir Akses Sistem Operasi: Larang akses langsung ke disk lokal atau socket jaringan kecuali yang diberikan izin secara eksplisit.

---

Jebakan Umum yang Harus Dihindari

Praktik Buruk	Dampak Buruk	Solusi Rekayasa
Menggunakan eval() untuk menjalankan kode AI	Risiko pelarian sandbox dan kompromi server host	Jalankan secara eksklusif di dalam sandbox WASM
Menjalankan migrasi langsung ke staging	Database corrupt dan downtime saat rollback gagal	Jalankan dry-run di WASM Postgres sebelum deploy
Membiarkan memori WASM tanpa batas	Crash server akibat memori habis (OOM)	Batasi jumlah halaman memori maksimum

---

Integrasi dengan Skill Lain (WAJIB)

• autonomous-tdd-debugger — Menggunakan sandbox sebagai lingkungan eksekusi pengujian mandiri.
• wasm-edge-computing-expert — Sinkronisasi profil runtime dan batasan memori WASM.
• autonomous-red-teamer — Menjalankan uji penetrasi dan fuzzing dalam lingkungan terisolasi.
• production-ready-hardener — Memverifikasi kebijakan keamanan sandbox dan alokasi resource.

Direferensikan oleh Orchestrator (WAJIB)

• brainstorming — Tambahkan ke baris "Testing & Keamanan" dan "Backend & Runtime" pada Matriks Orkestrasi.
• zero-to-prod-orchestrator — Tambahkan ke Fase 6 (Pengujian Otomatis, Ketahanan Error & Audit Keamanan).