This Claude Code Skill Built An 89K-Lead TAM In Minutes (Steal it)

If you build outbound for a living (or your team does), you know the ritual. Apollo seat. Clay credits. A spreadsheet. Two weeks of cleanup. A list that's stale before the first send.

The 2026 version is different. It's a Claude Code skill, three MCP servers, and a terminal. One command pulls your full TAM in minutes, deduped across every source, and hands it straight to your enrichment pipeline.

Last month I ran it on roofing contractors for a client. 113,000 raw records from 20+ sources. Deduped to 89,562 unique companies. 20,557 campaign-ready leads. Total cost: less than $15.

This email shows exactly how the skill works. At the bottom, you can grab the SKILL.md file and run the same command on your own market today.

~6 minute read

Want the /tam-map SKILL.md file? Reply "SEND IT" to this email and I'll drop it in your inbox today.

Replace Your Next Clay Or Apollo Build With A Terminal Command

What: Stop building TAM lists in Clay or buying them from Apollo. Build them in Claude Code instead, using a skill + MCP servers.

Why: Clay charges per enrichment and caps your throughput. Apollo gives you the same list every agency is already emailing - there's no edge in a commodity database. A terminal-based build runs in minutes, costs almost nothing, and pulls from sources most outbound teams never touch (state licensing boards, SAM.gov, trade association directories, manufacturer partner locators).

How:

Install the /tam-map skill (reply to this email for the file).
Drop it in ~/.claude/skills/tam-map/
Run: claude /tam-map "your ICP description in natural language"

10-20 minutes later you have three tiered CSVs (tier A, tier B, catch-all) ready to drop into EmailBison or your email sending platform of choice.

The 2026 Stack

Three MCP servers. Nine data tools. One skill to orchestrate them all.

MCP Servers (add to your ~/.claude/.mcp.json):

aiark - firmographic + technographic company discovery
discolike - natural-language lookalike search across 65M companies
spider - AI-ready web crawling for structured data from any site

Data sources the skill pulls from:

AI Ark - firmographic discovery (HIGH priority)
Discolike - lookalike expansion (HIGH)
Spider.cloud - tech + intent signals (HIGH)
Serper - Google Maps / Search API for local-business TAM (HIGH)
BlitzAPI - owner discovery + firmographics with unlimited credits (HIGH)
Prospeo - LinkedIn URL to email + mobile, 63% mobile hit rate on US professionals (MED)
IcyPeas - keyword-based company + people search (MED)
MillionVerifier - SMTP verification + catch-all flagging (HIGH)
Claude Haiku - LLM triage for fit scoring and summarization (HIGH)

The Walkthrough (Real Client, Real Numbers)

Here's what the skill actually did on a recent roofing TAM build.

Day 1: ICP + Install. Define the universe in natural language. Drop the /tam-map skill into ~/.claude/skills/tam-map/. Five minutes.

Day 2: First Scrape. Run one command. The skill kicks off Phase 1 (source discovery) and catalogs 16+ source categories: state contractor boards in TX, NRCA member directory, GAF and Owens Corning certified contractor locators, SAM.gov federal contract recipients, BBB, Google Maps via Serper ZIP scraping, Houzz, Thumbtack, Inc 5000 roofing companies, regional trade groups. 113,000 raw records pulled across sources.

Days 3-5: Enrich + Verify. The enrichment waterfall runs. BlitzAPI first (unlimited credits). Prospeo fallback for the misses. MillionVerifier does SMTP verification and flags catch-all domains into their own bucket. This matters. Agency-heavy TAMs bounce 35% on catch-alls at delivery even when MillionVerifier says they're valid.

Dedup + Segment. The skill deduplicates across every source using domain as the primary key, then falls back to phone number. Names drift. Domains don't. The 113K raw records collapsed to 89,562 unique companies.

Claude Haiku then scores each company 0-100 on fit and segments into three tiers plus the catch-all bucket. Output: tier-a.csv (20,557 leads), tier-b.csv, tier-c.csv, catch-all.csv. Every row EmailBison-ready.

Days 6-10: Ship. Upload the segmented CSVs straight into EmailBison. Tier-A goes to your sharpest copy. Tier-B gets the volume play. Catch-all runs on its own sender pool so the bounce doesn't kill your main domain rep.

Total cost for the full TAM: less than $15 in API credits. No Clay subscription. No Apollo seat. No per-lead fees.

The same skill works on any B2B vertical. SaaS, home services, e-commerce, legal, healthcare, construction, manufacturing. Change the ICP prompt, rerun the command, get a new tiered database.

Why This Matters

The agencies and in-house teams that will dominate outbound in 2026 aren't the ones with the best copy or the most senders. They're the ones whose TAM maps are closest to the actual market.

Flat TAM is why a team can send 100K emails a month and book 8 meetings. Survey-mapped TAM, built with this kind of terminal-first approach, is why a team can send 15K emails and book 45. Same category. Same budget. Different map.

The skill is free. The stack costs about $15 per full TAM build. The only real moat is knowing how to run it.

How I Can Help?

Get the /tam-map SKILL.md file: reply "SEND IT" to this email
If you run an agency, I want to know you: I refer deals to partners, feature builds in this newsletter, and swap clients when the fit is wrong. No fees, no contracts. Book a 15-min collab call: calendly.com/adam-revgrowth/collab-calls
Book a call for a free GTM strategy : calendly.com/adam-revgrowth/30min
Join Outbound Secrets (70+ training videos, code REV20 for 20% off): outbound-secrets.com
Reply to this email with questions - I read every one

Hope you found this useful. Reply "SEND IT" if you want the working SKILL.md file and I'll drop it in your inbox today.

Have a blessed day,

Adam