Prospect List Enrichment

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

1. Run python tools/batch_prospect_pipeline.py --input --product --market --tone --discovery --output-dir when one-step batch output is needed.
2. If the product file is a catalog and the user names one product, add --product-query or --sku instead of editing the catalog.
3. Read CSV or Excel rows without changing the original file.
4. 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.
5. Normalize websites by domain and deduplicate rows by domain first, then by normalized company name and country.
6. Preserve all source URLs and merge useful source notes when duplicates are combined.
7. Mark rows without company name or website as needs_review.
8. Mark rows outside the target market or unrelated to the product category as excluded_rows with a reason.
9. Write prospects.enriched.xlsx, research_reports.json, scores.xlsx, and email_drafts.xlsx when using the batch pipeline.
10. Treat research_reports.json as the source of fetched evidence for scoring and emails.
11. Do not fill missing company facts from assumptions; use needs_review, fetch_failed, or no_evidence.
12. 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.
13. 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 |

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Xuxchloris
• Source: Xuxchloris/hermes-export-skills
• License: MIT

Install and usage instructions live in the source repository linked above.