Prospect list enrichment

Skill Xuxchloris/export_skills/skills/prospect-list-enrichment

面向外贸出口场景的 Hermes / Open claw Agent Skills 包，支持海外客户发现、批量名单处理、公司背调、客户评分、决策层线索、开发信、跟进计划和报价单导出。

Install

npx -y skills add Xuxchloris/export_skills --skill prospect-list-enrichment

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

19 stars19 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when a trade agent needs to clean, deduplicate, normalize, enrich, or prepare CSV and Excel prospect lists before company research, scoring, or campaign planning

SKILL.md

3.8 KB, as published. Nobody here has run it

Prospect List Enrichment

Overview

Prepare CSV and Excel prospect lists for research and scoring. The core rule is to preserve source evidence while turning inconsistent rows into a reviewable company-level queue.

Do not create generated company rows. Every enriched row must come from the user's file or a configured collection output.

When to Use

Use this skill when the user provides a CSV or Excel customer list, asks to remove duplicates, wants missing fields flagged, or needs a batch queue for company research.

Use prospect-discovery first when the user needs a sourcing strategy instead of list processing.

Inputs

CSV or Excel prospect list
Product context from product-loader
Market rules from MARKET.yaml
Optional discovery output from prospect-discovery

Outputs

{
  "input_rows": 0,
  "unique_companies": 0,
  "duplicate_rows": [],
  "ready_for_research": [],
  "needs_review": [],
  "excluded_rows": [],
  "output_columns": [],
  "next_steps": ["company-research", "prospect-scoring"]
}

Procedure

Run python tools/batch_prospect_pipeline.py --input <prospects.csv|xlsx> --product <PRODUCT.yaml> --market <MARKET.yaml> --tone <TONE.yaml> --discovery <DISCOVERY.yaml> --output-dir <folder> when one-step batch output is needed.
If the product file is a catalog and the user names one product, add --product-query <name> or --sku <sku> instead of editing the catalog.
Read CSV or Excel rows without changing the original file.
Normalize column names and map available values to company name, website, country, business type, source URL, source note, contact email, contact phone, and contact clue.
Normalize websites by domain and deduplicate rows by domain first, then by normalized company name and country.
Preserve all source URLs and merge useful source notes when duplicates are combined.
Mark rows without company name or website as needs_review.
Mark rows outside the target market or unrelated to the product category as excluded_rows with a reason.
Write prospects.enriched.xlsx, research_reports.json, scores.xlsx, and email_drafts.xlsx when using the batch pipeline.
Treat research_reports.json as the source of fetched evidence for scoring and emails.
Do not fill missing company facts from assumptions; use needs_review, fetch_failed, or no_evidence.
If the user asks for a different output shape, preserve the same row data and choose the requested export format instead of changing the facts.
Send research-ready rows to company-research; send researched rows to prospect-scoring.

Verification

Original input file remains unchanged.
Batch output includes prospects.enriched.xlsx, research_reports.json, scores.xlsx, and email_drafts.xlsx.
Duplicate handling preserves source evidence.
Every ready row includes company name, website, and source URL.
Enriched rows include contact_email, contact_phone, email_result, and phone_result; missing contact values are written as 没有.
Missing fields are listed instead of guessed.
No generated rows appear in output workbooks.
Scoring and email drafts are based on fetched evidence.
Research happens before scoring.
Output format requests are honored without changing row facts.

Common Mistakes

Mistake	Fix
Deduplicating only by display name	Prefer normalized website domain
Dropping source URLs during merge	Preserve every useful source record
Treating missing fields as empty facts	Mark them for review
Scoring raw rows immediately	Run company research first