DigiVendAgent

Codename: Vendy — the sales & conversion agent of the team.

"Digi (digital) + Vend (sell) + Agent" — Vendy is step ⑤ of a six-agent pipeline. She takes a finished digital product and turns it into money: list → price → fulfill → collect → withdraw, looping until the target is met. (Research = Scout, production = Wright, build = Mason, traffic = Buzz, analysis = Echo.)

Vendy's objective is the sales side of passive income. She doesn't take hourly freelance work or offer courses / managed services / consulting. Given a "produce-once, sell-many" digital product, she autonomously lists, prices, fulfills, collects payment, handles after-sales, and withdraws.

简体中文

---

🎯 Positioning

Dimension	Detail
What	Autonomously sell digital goods online for profit (team step ⑤ — sales / conversion)
What not	No time-billed freelance; no human-intensive active services
Market	Global (English-first) + China (Xiaohongshu)
Payout	Withdraw to HK / CN bank (Payloadz + PayPal main route)
Mode	Continuous trial loop until the target is met

---

🧠 Core Capabilities

🔎 Trend research (/hot-trend) now lives in Scout — ProductStrategistAgent. Vendy is the conversion / sales agent: give her a product and she turns it into money.

Vending Engine · /vend (the sales loop)

Vendy's "hands". Reads the target from docs/config.json and runs the sales loop: list → price → fulfill → collect → check metrics → adjust. Producing the product is Wright's job (ProductProducerAgent); driving traffic is Buzz's (GrowthMarketerAgent). She only stops for three special cases; otherwise she runs autonomously until the target is met.

• Trigger: /vend
• Listing & payment: Payloadz → PayPal (overseas); Xiaohongshu (China)
• Fulfillment & after-sales: delivery, refunds, disputes
• Three cases that require stopping to ask: account registration / permissions, password / verification input, spending money

---

🔁 Typical Workflow

Vendy is step ⑤ of a six-agent team:

① Scout (research) → ② Wright (produce) → ③ Mason (build site) → ④ Buzz (traffic) → ⑤ Vendy (sell / ops) → ⑥ Echo (analyze)

Run the sales loop:

/vend   →   list → price → promote → collect → withdraw (loop until target met)

---

📦 What's in This Repo

This open-source repo ships the agent's design and skills — not runtime data.

DigiVendAgent/
├── README.md                 ← this file (English)
├── README_cn.md              ← Chinese README
├── LICENSE.md                ← MIT
├── CLAUDE.md / CODEBUDDY.md  ← project-level agent instructions (CODEBUDDY.md → CLAUDE.md)
├── CHANGELOG.md              ← version history
├── assets/                   ← logo and visual assets
└── .codebuddy/skills/        ← agent skills (vend, etc.)

Runtime data (products, scripts, logs, config.json with credentials) lives in docs/, which is gitignored and not distributed. Each user keeps their own locally.

---

⚙️ Configuration (runtime, local only)

Runtime data is centralized under docs/. Core config: docs/config.json.

⚠️ docs/config.json contains plaintext credentials / API keys. It is excluded by .gitignore and never committed. Structure only (values placeholder):

{
  "target":   { "amount": 30000, "currency": "HKD", "accept_crypto": true, "withdrawable_to": ["HK_bank", "CN_bank"] },
  "speed":    { "amount": 200, "currency": "HKD", "period": "hour" },
  "budget":   { "amount": 100, "currency": "HKD", "require_roi_before_spend": true },
  "assets":   { "product_dir": "docs/product", "history_dir": "docs/history", "log_dir": "docs/log" },
  "accounts": {
    "payloadz":   { "email": "", "password": "" },
    "PayPal":     { "email": "", "password": "" },
    "twitter":    { "email": "", "username": "", "password": "", "api_key": "", "api_secret": "", "access_token": "", "access_token_secret": "" },
    "instagram":  { "username": "", "email": "", "password": "" },
    "youtube":    { "channel_id": "", "email": "", "password": "" },
    "xiaohongshu":{ "username": "", "email": "", "password": "" }
  },
  "llm_api_keys": { "openai": "", "anthropic": "", "gemini": "", "openrouter": "", "Agnes": "", "devto_api_key": "" }
}

• target: money target (amount, currency, crypto acceptance, withdrawable accounts)
• speed: earning-speed target (internal efficiency KPI, not hourly freelance)
• budget: budget (require ROI estimate before spending)
• assets: local asset directories
• accounts: platform credentials
• llm_api_keys: LLM API keys

---

🚀 Quick Start

1. Create docs/config.json locally with your target, budget, and platform accounts (never committed).
2. (Optional) Get a validated product opportunity from Scout — ProductStrategistAgent — or supply your own product.
3. Start the sales loop:

   /vend
   

4. Vendy only stops for three reasons: account registration, password / verification, or spending money. Otherwise she runs autonomously until the target is met.

---

🛡️ Safety & Boundaries

• Only lawful, compliant money-making attempts; no exploits, no fake orders, no fraud.
• Credentials live only in local config.json; never written back to skills or logs.
• Never spends money without explicit user consent; every spend carries an ROI estimate.
• No traffic from politics, disasters, tragedies, or heavy controversy; no copyrighted / trademarked material.
• Earnings estimates are always given as ranges with uncertainty noted; never "guaranteed X per month".

---

📝 Design Principles

• Passive income first: produce once, sell many.
• Speed first: take the fastest legal path to revenue (internal KPI = 200 HKD/hour).
• Keep trying: if one channel fails, immediately move to the next.
• Reuse existing assets: consume inventory and lessons in product/ and history/ first.
• Reuse skills, don't reinvent: actively /find-skill, /install-skill.

---

📄 License & Attribution

MIT — see LICENSE.md.

Copyright (c) 2026 All Contributors.

If you use, fork, or build on this project, please credit DigiVendAgent (Vendy) and link back to the project repository: https://github.com/xhqing/DigiVendAgent.

Vendy keeps running until the money is made.