Introduction

Common approaches to foreign trade customer acquisition usually involve one of two methods: either throwing links from customs data, yellow pages, or maps to a general LLM to “search casually,” or dumping lists into a tool that only sends template emails. The former results in inconsistent outputs every time, while the latter lacks demand discovery, leaving customers permanently stuck on a list.

The foreign-trade-customer-dev introduced below is a community Skill on SkillHub (located at @user_5cafce6a/foreign-trade-customer-dev, version 1.0.0). It is positioned as a central hub for foreign trade customer development: connecting “Acquisition → Triage → Demand Discovery → Outreach Guidance” into a processing pipeline, with the goal of maximizing the sustainable, compliant, and continuous output of clients.

What is This

In short: It is not a search engine, nor is it a mass email sender. It is a customer development workflow that operates on a predefined pipeline, coordinated internally by multiple sub-processes.

  • Publisher: SkillHub Community Publisher user_5cafce6a
  • Category: business-ops
  • Requirements: requires_api_key=true. At least a search API Key is required to run the core acquisition pipeline.
  • Trigger Method: Invoked via Chinese or English trigger words (see “Typical Triggers” below).

Processing Pipeline: Four-Stage Model

The entire pipeline follows a sequential flow, where the output of one stage becomes the input for the next:

  1. Acquisition: Select a batch of sources from the channel matrix to pull raw leads (companies, contacts, contact info, source channels).
  2. Triage: Triage by scale (Large/Medium/Small) and budget (Zero/Small/Enterprise) to determine pipeline parameters for each lead.
  3. Demand Discovery: Perform deep background checks on leads to extract memory points such as main business, current products, and product needs.
  4. Outreach Guidance: Generate customized outreach material based on memory points and contact information.

Scale Triage and Budget

Users first select customer scale (Large / Medium / Small), then select the budget tier (Zero / Small / Enterprise). These two choices link to pipeline parameters:

  • Channel scope
  • Decision-maker mining depth
  • Background check depth
  • Outreach methods
  • Batch volume

This means that under the same trigger word, the acquisition and research intensity for a large client differs from that for a zero-budget lead. This avoids over-investing in small leads or running only lightweight processes for large clients.

Channel Matrix

Group A (Active Scraping) contains 22 categories; Group B (Passive Inquiry) contains 3 categories. For complete fields, see references/channels.md.

Group A (Active Scraping, 22 Categories)

  1. Customs Data
  2. Industry Yellow Pages
  3. Logistics Bills of Lading
  4. Google Maps
  5. LinkedIn
  6. Google Advanced Search (including government tenders, B2C top sellers)
  7. Local Vertical Search (Yandex / Naver / Seznam / Qwant / Baidu Thai / CocCoc)
  8. Local Private Domains (VK / Zalo / Telegram / LINE / WhatsApp)
  9. Overseas Forums & Q&A
  10. Exhibitions, Chambers of Commerce & Associations
  11. Patent Databases
  12. Wikipedia Entries
  13. Industry Report Repositories
  14. Offline Digital Directories
  15. Supply Chain Reverse Mining
  16. Global Tendering & Bidding (UNGM / World Bank / TED / SAM.gov / Etimad / Aggregators)
  17. Construction Associations & Chambers (AGC / NAHB / CIOB / SaudiBuild / SACA)
  18. Construction Materials Vertical B2B (Restado / TradeKey / Archspex)
  19. Resource Exchange (Freight forwarders / Certification / Overseas warehouses / Peer complement, manual entry)
  20. Community Reverse (Forums / YouTube Bloggers / Private Groups)
  21. Real Estate & EPC Design Institutes (Listed real estate firms / International General Contractors / Design Firms)
  22. Global Video Platforms (YouTube / Meta Reels / TikTok / Local Video / Industrial Verticals)

Group B (Passive Inquiry, 3 Categories)

  1. Lead submissions from independent websites
  2. Contributions to industry media
  3. Cross-border B2C top sellers

Search Cost: Three-Layer Ladder Scheduling

Search cost in the acquisition phase is the most detailed part of this Skill’s design. When --provider auto|ladder is enabled, searches are scheduled across four tiers based on cost, from low to high:

Tier0 Local OpenSERP ($0)
  → Tier1 Serper Multi-Key Polling (2500 free queries per key)
    → Tier2 Searlo Pay-as-you-go ($0.3/1000 queries)
      → Tier3 Alert/Backup

Switching rules:

  1. Tier0 Local OpenSERP is rate-limited to 4~6 seconds/request; 5 consecutive errors or CAPTCHA appearance triggers an automatic upgrade to Tier1.
  2. Tier1 polls multiple keys; when a single key’s quota is exhausted, it automatically switches to the next; if all are exhausted, it upgrades to Tier2.
  3. When Tier2 reaches the monthly budget cap, it upgrades to Tier3 for alerts.

State is persisted in ladder_state.json, recording monthly consumption, key polling, budget, and consecutive error counts, with automatic resets across months.

Two accompanying parameters:

  • --mega: Cross-engine aggregation and deduplication for anti-blocking.
  • --extract: Directly scrape landing page content to feed into Deep Research, while conveniently parsing contact info (email / WhatsApp / phone) and website characteristics.

Note: OpenSERP does not support Maps and Video queries. Such queries will automatically start from Tier1 because Serper natively supports Places and Videos.

Contact Info and Website Profile Extraction

This is the raw material for outreach emails. Each lead automatically completes four types of fields:

  • email: Enterprise business emails only; personal gmail/qq etc. are excluded.
  • whatsapp
  • phone
  • website_profile: Certification status, positioning, market, year established.

Two acquisition paths:

  1. --extract: Captured cost-free from OpenSERP landing page content during acquisition.
  2. --enrich (or standalone --enrich-file <csv>): Connect directly to the official website after acquisition, with OpenSERP as a fallback for completion.

Compliance Boundary: Only public enterprise contact information is collected, adhering to GDPR. Personal emails and mobile numbers are not scraped.

Memory Point Collection and Two-Phase Backfill

The first round of API batch searches (Serper / Searlo / OpenSERP) extracts memory points for each lead in addition to names, addresses, and contact info:

  • memory_main_business: Main business
  • memory_products: Products on sale
  • memory_product_needs: Product needs

These fields are fed directly into subsequent customized outreach emails. Missing fields are marked needs_local_fill=True and are filled in by a second-phase local separate search:

--local-fill-file <file>

The second phase involves targeted searches or direct connections to official websites, without relying on fabrication. Each lead also records search_term and source_channel, which are accumulated to optimize search strategies: adding terms for high-quality queries and down-weighting low-quality channels.

Memory Principles Document and Industry-Country Classification

Upon first launch, the Skill detects if “Memory Principles.md” exists in the workspace:

  • Missing: Initializes from the template references/memory-principles.md.
  • Then guides the user to fill in or supplement it, providing optimization suggestions after review (e.g., if the plan is too thin or principles are too rigid).

Customer information is archived by default under industry-country tags, such as diaper-japan, cnc-usa. API Keys, PATs, and other critical information are centrally registered in this document; sensitive credentials are only recorded by name and storage location, not in plain text.

Deep Background Check (Deep Research)

Deep background checking is the core loop turning lists into insights, taking priority over other search capabilities. Research loops are run in tiers based on scale:

  • Large Clients: 7-step complete research
  • Medium Clients: 4~5 steps
  • Small Clients: Lightweight research

It produces a dossier with cited sources and incrementally updates the CRM signal library to avoid re-researching every time. See references/deep-research.md for details.

Compliance and Sustainable Growth

In terms of compliance:

  • EU opt-out channel
  • Daily send limit per domain ≤ 150 emails
  • Outreach emails automatically match three versions based on scale

In terms of sustainable growth, a set of mechanisms ensures customer acquisition does not run dry:

  1. Matrix rotation
  2. CRM deduplication
  3. Second-degree network
  4. Supply chain reverse fission
  5. Signal-driven approach
  6. Passive inbound
  7. Channel priority weighting
  8. Scheduled scripts
  9. Compliance anti-exhaustion

Typical Triggers

Invoked via natural language trigger words rather than command-line calls. Chinese triggers cover a wide range, from acquisition channels to outreach:

  • Customer development / Find clients / Foreign trade acquisition / Outreach emails / Customer profiling / Deep background check / Customer background check
  • Customs data client finding / Industry yellow pages / Google Maps development / LinkedIn development / Local engine development / Exhibition directories
  • Bidding development / UN procurement / World Bank tenders / Tender client finding / Infrastructure & construction materials acquisition
  • Real estate developer client finding / EPC general contractor development / Design institute client finding / Forum reverse acquisition
  • Video platform development / YouTube client finding / TikTok development / Short video acquisition / Industrial video acquisition
  • Self-hosted search / OpenSERP / SerpAPI alternative / Free SERP / Acquisition cost optimization
  • Contact mining / Email mining / WhatsApp development / Website profile extraction

English trigger words include lead generation, outbound, deep research, sustainable acquisition, etc.

Built-in Sub-processes

This Skill internally calls six sub-processes:

  1. Deep Research Loop: The main engine from list to insight.
  2. ultimate-b2b-outreach: Discovery + Mining.
  3. b2b-lead-research: Decision makers.
  4. b2b-deep-diligence: Formal Word background report for large clients.
  5. crm-manager-skill: Pipeline accumulation.
  6. linkedin-lead-monitor: Signal monitoring.

Use Cases and Limitations

Suitable for: Foreign trade teams or independent salespeople who need long-term, continuous, multi-channel customer acquisition, especially for scenarios with high unit prices, long decision chains, and the need for customized outreach emails.

Limitations and Pitfalls:

  1. Requires a Search API Key (requires_api_key=true). The zero-budget path relies on local OpenSERP, but Maps and Video queries must go through Serper; this cost cannot be saved.
  2. Deep background checks rely on internet-based research capabilities; in restricted network environments, only lightweight processes can be run.
  3. The three stages of acquisition, enrichment, and local-fill require official website accessibility; if sites are blocked or have anti-crawler measures, it degrades to OpenSERP fallback.
  4. This Skill targets B2B foreign trade scenarios; B2C top sellers are only used as sources for passive inquiries and reverse leads, not the main battlefield.
  5. The public page crawled this time did not return the main text. Installation commands and configuration fields should be based on the official SkillHub page; this article does not fabricate them.

Conclusion

The value of this Skill lies in transforming foreign trade customer acquisition from “finding a batch of emails and sending them” into a processing pipeline with triage, cost ladders, compliance boundaries, and memory accumulation. First, run the channel matrix to get leads, then triage by scale to determine investment depth, and finally use background checks and memory points to support customized outreach.

Official Address: https://api.skillhub.cn/user_5cafce6a/foreign-trade-customer-dev
SkillHub Page: https://www.skillhub.cn/skills/user_5cafce6a/foreign-trade-customer-dev