GetestEnGoed Review Pipeline

5 AI bots. 1 honest review. ~25 seconds.

A production-ready Claude Code skill for generating data-driven product reviews using a team of 5 specialized AI bots. Built by GetestEnGoed.nl — a Dutch review platform where 5 AI bots form the editorial team, supervised by a human editor-in-chief.

---

The 5 Bots

Bot	Model	Role
🔍 Scout	GPT-4.1 mini	Finds product, specs, 3-5 competitors
📊 Data	GPT-4.1 mini	Spec comparison, price-per-feature score
🎙️ Voxy	GPT-4.1	Analyzes hundreds of real user reviews
💰 Penny	GPT-4.1 mini	Price analysis, TCO, buy advice
⚖️ Judge	GPT-5.1	Final synthesis + score (0-10)

Cost per review: ~$0.22-0.25 | Generation time: ~15-30 seconds

---

How It Works

Product Name / EAN / URL
        ↓
   Scout (sequential)
        ↓
   Data + Voxy + Penny (parallel)
        ↓
   Judge (final synthesis)
        ↓
   Published Review

The score is composed of 4 weighted criteria:

• User satisfaction (Voxy): 35%
• Specs vs. competition (Data): 25%
• Price-to-value (Penny): 25%
• Availability & ecosystem (Scout): 15%

---

Installation

git clone https://github.com/yeah1966/getestengoed-review-pipeline.git
cd getestengoed-review-pipeline
bash install.sh

Requirements:

• Claude Code
• Node.js ≥ 18
• OPENAI_API_KEY in your environment

---

Usage

In Claude Code:

/review Dyson V15 Detect          — Full review
/review 8720683438521             — Review by EAN barcode
/review plan                      — Explain the pipeline
/review bots                      — List all bots and roles
/review score <product>           — Show scoring breakdown

---

What You Get

Each review includes:

• Editorial title — honest, provocative (max 70 chars)
• Score — 0-10 with 1 decimal
• Hero description — 2-3 sentences with target audience
• 5 bot sections — scout/data/voxy/penny/conclusion
• Comparison table — product vs 3-5 real competitors
• Pros & cons — concrete, no marketing speak
• Suitable for / not suitable for — explicit user profiles
• FAQs — min 4 real shopping questions

---

Key Rules

This pipeline enforces strict honesty rules:

1. No fake testing claims — bots analyze data, they don't physically test products
2. Prices only from verified API sources — never invented
3. No marketing patterns — "Nowadays...", "In the current market..." are blocked
4. Conflict detection — when user satisfaction and spec scores diverge ≥2 points, Judge must address it explicitly

---

Live Example

Every review on GetestEnGoed.nl is generated by this exact pipeline. Browse the site to see the output in production.

---

Files

├── SKILL.md                          # Claude Code skill definition
├── install.sh                        # Installation script
├── uninstall.sh                      # Uninstall script
└── references/
    ├── pipeline-architecture.md      # Full bot details + data flow
    ├── bot-outputs.md                # TypeScript output types
    └── scoring-model.md              # How scores are calculated

---

Author

Built by Ron Bode — founder of GetestEnGoed.nl

• LinkedIn: linkedin.com/in/ron-bode-3629597b
• GitHub: github.com/yeah1966

---

License

MIT — use freely, attribution appreciated.