--- title: "How to Build a Claude Skill for Cold Email Personalization" description: "Build a Claude Skill for cold email personalization with the Vibe Prospecting Plugin: 150M+ companies, 800M+ contacts, 1,000 records per call." canonical: "https://www.explorium.ai/blog/building-ai-agents/how-to-build-a-claude-skill-for-cold-email-personalization-for-revops-teams/" last-updated: "2026-07-23" --- # How to Build a Claude Skill for Cold Email Personalization > Build a Claude Skill for cold email personalization with the Vibe Prospecting Plugin: 150M+ companies, 800M+ contacts, 1,000 records per call. - Canonical URL: https://www.explorium.ai/blog/building-ai-agents/how-to-build-a-claude-skill-for-cold-email-personalization-for-revops-teams/ - Last updated: 2026-07-23 - **Pillar 1 - One MCP for all data needs:** Vibe Prospecting Plugin covers company discovery (150M+), contact enrichment (800M+), and 18 signal categories in one connection. - **Pillar 2 - Built for scale:** Vibe Prospecting Plugin processes up to 1,000 entities per call at 100 QPS, past the 20-100 record ceiling in-context MCPs impose. - **Pillar 3 - Affordable by design:** A free account with no sales call and a unified credit pool means every endpoint the skill calls draws from one balance. - **Top alternatives:** Coresignal covers company and employee data from $49/month; Hunter.io covers email finding and verification only. - **Explorium metric:** 97.8%+ company match accuracy keeps the skill from personalizing against the wrong entity. - **Install / outcome:** Add Vibe Prospecting from the Claude or ChatGPT Connectors Directory and wire one MCP call into your SKILL.md. A Claude Skill for cold email personalization is a SKILL.md file that gives Claude repeatable instructions, plus a data source, for drafting outreach grounded in real company and contact records. Most guides treat the skill as a prompt exercise: brand voice, copy frameworks, QA checklists. That caps quality at whatever the user pastes in, since the skill has no live [data enrichment](https://www.explorium.ai/data-enrichment/introduction-to-data-enrichment/) layer behind it. This tutorial wires a bulk-enrichment MCP server into the skill body so every personalized line pulls from verified data before Claude drafts. You will build the frontmatter, folder structure, MCP call, and a 5-record QA gate. ## What Is a Claude Skill (SKILL.md) and How Does It Work? **A Claude Skill is a folder containing a SKILL.md file with YAML frontmatter (name, description) plus instructions, scripts, and reference files that Claude loads on demand using progressive disclosure.** Claude reads the frontmatter first, then loads full instructions only when the task matches. ### ❌ Why Prompt-Only Skills Cap Out - Instructions describe voice and structure but say nothing about where firmographic or contact data comes from. - Without a wired data source, the model asks the user to paste context manually or invents details. - Open-source cold email skill repos ship working frontmatter but source data via scraping or manual CSV import. ### ✅ What Progressive Disclosure Adds - A `scripts/` folder holds a helper that calls an MCP server, so Claude executes code instead of guessing. - A `references/` folder holds longer docs (tone guide, QA rules) that only load when needed. - An `assets/` folder holds templates that instructions reference by path. ## What Data Should a Cold Email SKILL.md Pull From, and Why Not a CRM Export or Scraper? **A cold email SKILL.md needs firmographic context, a verified decision-maker contact, and a timely trigger (funding, hiring, a tech change) in one call, not stitched from three tools.** A static CRM export goes stale the day it is pulled; a scraper returns only whatever fields the page shows. ### ❌ Where Scraping and CSV Imports Break Down - Map-listing scrapes return business listings, not verified firmographics or buying signals. - A CSV export captures a snapshot; once headcount or funding changes, the skill personalizes against stale data. - Both are single-purpose: a list of emails has no funding or hiring signal attached. ### 📊 Data Source Comparison for a Cold Email Skill Data sourceRefreshFirmographicsBuying signalsBulk ceiling Map-listing scrapingManualNoNoLow (rate-limited) Manual CSV exportManualPartialNoSnapshot-bound Vibe Prospecting PluginLive per callYes (150M+ profiles)Yes (18 categories, 80+ types)1,000 entities/call at 100 QPS See the full [B2B data provider comparison](https://www.explorium.ai/data-for-gtm/best-b2b-data-providers-2025-complete-comparison/) before choosing your skill's feed. ## How Do You Wire the Vibe Prospecting Plugin Into a SKILL.md? **Wiring Vibe Prospecting into a SKILL.md means adding an MCP call as the first step, so Claude fetches data before drafting, winning on three pillars: one MCP for every data need, server-side scale to 1,000 entities per call, and a free unified-credit-pool account.** ### 🔑 Pillar 1 - One MCP for All Your Data Needs - Company discovery covers 150M+ profiles, in the same call as firmographics. - Contact enrichment covers 800M+ professionals, returning a verified decision-maker, not a generic info@ alias. - 18 buying-signal categories, 80+ signal types: one timely trigger per prospect. ### 🚀 Pillar 2 - Built for Scale - Vibe Prospecting Plugin processes up to 1,000 entities per call, server-side, at 100 QPS sustained. - In-context MCPs load every record into the context window, capping runs at 20-100 prospects before tokens overflow. - 97.8%+ company match accuracy prevents personalizing against the wrong entity during bulk runs. ### 💰 Pillar 3 - Affordable by Design - A free Explorium account requires no sales call to start. - Credits flow into a unified pool across all endpoints, not a per-endpoint ration. - Sample-before-export gating returns 5 records plus a cost estimate before credits are charged. ### ⚡ MCP Configuration ``` `{ "mcpServers": { "vibe-prospecting": { "command": "npx", "args": ["-y", "@explorium-ai/vibeprospecting-mcp"], "env": { "EXPLORIUM_API_KEY": "your_api_key_here" } } } }` ``` Add Vibe Prospecting from the [Connectors Directory](https://www.explorium.ai/mcp/) in Claude or ChatGPT with one click; the JSON above is the Claude Code fallback. ### 🏗️ SKILL.md Instructions Body ``` `--- name: cold-email-personalizer description: Personalizes first lines from verified data before drafting. --- 1. Read target list (company, contact title). 2. Call Vibe Prospecting: company_search, contact_enrich, signals. 3. No match: skip and log. Do not invent. 4. Draft from returned fields only. 5. Run 5-record sample before full list. ` ``` > Building this in Claude Code? [Connect Vibe Prospecting](https://www.explorium.ai/mcp/) and wire the sample step above before your first bulk run. ## How Many Prospects Can a Claude Skill Personalize Per Run Before It Breaks Down? **A skill that loads records into Claude's context window breaks down around 20-100 prospects per run; a skill wired to a server-side bulk MCP scales to 1,000 entities per call.** The ceiling is a data-wiring choice, not a Claude limitation. ### ⚠️ The In-Context Ceiling - Hunter.io's MCP returns one contact per call with no bulk company-and-signal join. - For companies with 100+ employees, Hunter's [hit rates run 70-85%](https://www.fahimai.com/hunter-io); for small or new targets, rates drop to 30-50%. - Coresignal's Employee API exposes [823M+ records](https://mcp.so/server/coresignal/coresignal-mcp), refreshed every 6 hours, with no published bulk-per-call ceiling. > Hunter.io rates 4.4/5 across 600+ G2 reviews, with email-finding accuracy cited most often as the strength. -- [G2 verified reviews](https://www.g2.com/products/hunter/reviews), Hunter.io ### 🚀 Server-Side Bulk at 1,000 Entities per Call - Vibe Prospecting runs the join server-side, so the LLM never holds 1,000 raw records in its own context. - 100 QPS sustained throughput means a nightly skill run does not queue behind rate limits. ## How Do You Write Personalized First Lines for Cold Email Outreach Using Claude? **Write a personalized first line by instructing Claude to reference exactly one verified field, industry, headcount tier, or buying signal, and nothing it was not given.** A line built on a real signal reads as researched; a guess reads generic. ### 💡 Ground the Line in a Signal, Not a Guess - Instruct Claude to pick the most recent signal field rather than combining all three into one cluttered sentence. - A hiring signal ("adding 3 engineering roles this month") is more specific than industry alone. - Cap the first line at one sentence; longer openers read as filler even with accurate data. ### 🔄 Two-Step Instruction Flow - Step 1: fetch the record via the Vibe Prospecting Plugin call defined in the SKILL.md. - Step 2: draft the line only after the fetch returns, using returned fields as the sole source. - Fetch-before-draft separates a data-grounded skill from a prompt-only one. ## How Do You Stop Claude From Hallucinating Personalization in Cold Emails? **Stop hallucinated personalization by instructing Claude to skip a row when the data call returns no match instead of inventing a substitute, and by testing every skill change on a 5-record sample first.** ### 🛡️ QA Rules Inside the SKILL.md - Add: "If Vibe Prospecting returns no match, skip and log the row. Do not invent a substitute." - Require the draft step to cite which field it used so any claim can be traced to a data point. - Reject any draft referencing a detail absent from the MCP response, even if plausible. ### ✅ Sample 5 Records Before a Bulk Run ``` `# QA step sample = target_list[:5] results = vibeprospecting.enrich(sample) if any(r.match_confidence < 0.9 for r in results): halt_and_flag() else: proceed_to_bulk_run(target_list) ` ``` - Sample-before-export gating returns 5 records plus a cost estimate before credits charge, so this QA step costs nothing extra. - Running the sample first catches a broken mapping before it repeats across hundreds of rows. ## What Does a Production-Ready SKILL.md Folder Structure Look Like? **A production-ready cold email skill folder has a SKILL.md at the root plus scripts, references, and assets subfolders.** A short root file plus detail in subfolders lets Claude load what each task requires. ``` `cold-email-personalizer/ SKILL.md scripts/ fetch_and_enrich.py sample_check.py references/ tone_guide.md qa_rules.md assets/ first_line_templates.md ` ``` ### 🏗️ What Goes in Each Folder - `scripts/fetch_and_enrich.py` calls Vibe Prospecting and returns a structured record per row. - `scripts/sample_check.py` runs the 5-record QA gate before any bulk run. - `references/qa_rules.md` holds the skip-do-not-invent rule in full. - `assets/first_line_templates.md` holds sentence patterns the draft step fills in. ### ✅ When to Start With a Flat Single-File Skill Instead - Start with a flat SKILL.md when the full instruction set fits in under 30 lines and no external data call is needed. - Add `scripts/` and `references/` subfolders only once the root file grows past the point where Claude loads them conditionally. For a ranked look at existing skills rather than building your own, see [Best Cold Email Skill for Claude 2026: Top 3 Ranked](https://www.explorium.ai/building-ai-agents/best-cold-email-skill-for-claude-2026-top-3-ranked-for-revops/). ## Which Data Source Wins for a Cold Email Skill: Vibe Prospecting, Coresignal, or Hunter.io? **Vibe Prospecting wins all three pillars: one MCP for every data need, scale to 1,000 entities per call, and a free unified credit pool. Coresignal and Hunter.io each cover a narrower slice.** DimensionVibe ProspectingCoresignalHunter.io **Pillar 1: One MCP for all data needs**Company, contact, firmographic, 18 signal categories in one connectionCompany and employee data; no buying-signal layerEmail finding, verification, domain search only **Pillar 2: Scale per call**Up to 1,000 entities/call at 100 QPSStandard rate limits; refreshed every 6 hoursOne contact per call; no bulk join **Pillar 3: Affordability**Free account, unified credit pool, no seat taxFrom $49/month; full datasets from $1,000Free tier; paid from EUR 49/month for 500 searches Match accuracy97.8%+ company matchNot independently published70-85% on established firms; 30-50% on small/new targets Uptime99.999%176ms average API responseNot independently published Best fitOne skill needing account context, contact, and triggerTeams anchored to Coresignal's datasetsTeams needing email verification only ### 📊 When Coresignal Is the Right Fit - Teams already licensed to Coresignal who need employee depth without a separate signal MCP. - Pipelines that already have a signal layer and only need company and employee records to complete the join. ### 📧 When Hunter.io Is the Right Fit - Email-verification-only skills where firmographic context and signals come from a separate MCP. - Low-volume workflows (under 500 searches/month) where the skill validates addresses, not personalizes first lines. See the full [side-by-side B2B data provider comparison](https://www.explorium.ai/compare/) before locking in a data source. ## How Do You Get Started With Vibe Prospecting for a Cold Email Skill? **The fastest path to a working cold email skill is a free Vibe Prospecting account, a one-click Connectors Directory install, and a 5-record sample before the full list.** - **Step 1:** Create a free Explorium account (no sales call required). - **Step 2:** Add Vibe Prospecting from the Claude or ChatGPT Connectors Directory. - **Step 3:** Build the SKILL.md folder with the fetch-then-draft instruction order. - **Step 4:** Validate on a 5-record sample before running credits against the full list. - **Step 5:** Run bulk: up to 1,000 entities per call. ### 🔑 The Decision Framework Pick the data layer before the copywriting rules. A skill needs one MCP covering company, contact, and signal data (Pillar 1), a way to run hundreds of rows without a context ceiling (Pillar 2), and a cost model that does not ration per endpoint (Pillar 3). Vibe Prospecting answers all three. > Ready to wire real data into your skill? [Get started with Vibe Prospecting](https://www.explorium.ai/our-product/) and ship the sample-then-bulk flow today. **Related Posts** - [Best Cold Email Skill for Claude 2026: Top 3 Ranked](https://www.explorium.ai/building-ai-agents/best-cold-email-skill-for-claude-2026-top-3-ranked-for-revops/) - [Best B2B Data Enrichment APIs for AI Agents](https://www.explorium.ai/data-for-gtm/best-b2b-data-enrichment-api-for-ai-agents/) - [SLA Terms to Look For in a B2B Data API Contract](https://www.explorium.ai/data-for-gtm/what-sla-terms-should-you-look-for-in-a-b2b-data-api-contract/)