The Claude Code GTM Playbook — 8 Prompts to Run Your Outbound Stack

This is a working operator's asset. Every prompt below is meant to be copy-pasted into a Claude Code session and run today. If you have an MCP installed and a CSV on disk, you can execute the entire outbound stack — sourcing, enrichment, scoring, CRM hygiene, follow-ups, copy, and campaign launch — without touching Zapier, n8n, Clay, or a single engineer.

Each section gives you the operation, the setup, the exact prompt, and one tactical note most operators miss.

Section 01Browser Scraping

OperationScrape any company website for prospect intel before you write a single line of copy.

Generic personalization tokens (first name, company) are dead. What actually moves reply rate is a one-line opener that proves you read their site. Claude Code's browser tool lets you turn any URL into a structured intel file — founders, stack, news, hiring, pricing tells — in under a minute per account, no scraping infrastructure required.

Prerequisite setupClaude Code with browser/web tools enabled (no MCP required). Have a list of company domains ready in a text file or paste them inline.

Use your browser tool to visit each of the following company URLs and extract a structured intel report. For each company, return a markdown block with these exact fields: - Company: - Domain: - Founders / Leadership Team: (names + titles, pulled from /about, /team, or LinkedIn references on the site) - Tech Stack Signals: (any tools mentioned — CRM, ESP, analytics, sales platform — pulled from job posts, blog, or footer integrations) - Recent News / Announcements: (last 90 days only — funding, product launches, hires, partnerships) - Open Roles: (titles + departments from /careers or /jobs) - Pricing Page Signals: (tier names, price points, "contact sales" gates, free trial vs demo CTAs, target segment language like "for teams" vs "enterprise") - One-line Outbound Hook: (the single most personalizable detail from the above, written as a sentence I could open a cold email with) URLs: [paste 1-20 domains here, one per line] Output as a single markdown file I can save to disk.

Pro tipRun this in batches of 10–20 domains and append a column for "outbound hook" — the one specific sentence you'd lead a cold email with. That column is the entire ROI of this prompt.

Section 02Buying Signal Detection

OperationRead a company's jobs page to detect whether they are actually in-market for what you sell.

A company hiring 4 SDRs and 1 AE is building outbound capacity right now. A company hiring 1 RevOps lead is fixing a stack. A company posting "experience with HubSpot, Apollo, Outreach" in a job description has just told you their entire tech stack for free. Job pages are the highest-signal, lowest-cost intent dataset on the open web.

Prerequisite setupClaude Code with browser tools enabled. Have the company's careers/jobs URL (usually /careers, /jobs, or a Greenhouse/Lever subdomain).

Visit the following company careers pages. For each company, extract every open role and return a structured analysis: For each company output: - Company: - Total open roles: - Sales-org roles only: (list each — title, level, department) - SDR count: - AE count: - Sales leadership count: (VP Sales, Head of Sales, CRO, Director of Sales) - SDR-to-AE ratio: - Tools explicitly named in any sales job description: (CRM, sequencer, intent data, dialer, enrichment — quote the exact line they appear in) - Hiring velocity signal: (are they backfilling, scaling, or building from zero? Infer from role mix and seniority) - Buying-intent verdict: (HOT / WARM / COLD — and one sentence of reasoning) Careers URLs: [paste URLs here] Sort the final output by buying-intent verdict, HOT first.

Pro tipCompanies hiring 3+ SDRs simultaneously are nearly always evaluating new outbound tools in the same quarter. That window is 30–60 days. Hit them now.

Section 03Lead Enrichment at Scale

OperationEnrich a list of names + company domains with verified, deliverable emails.

Bounce rate kills sender reputation faster than bad copy. Findymail returns verified emails with a confidence score, and the Findymail MCP lets Claude run the enrichment loop directly — no CSV uploads to a web app, no copy-paste workflow, no cost on bad guesses. The MCP only charges on verified hits.

Prerequisite setupFindymail MCP installed in Claude Code (configured with your Findymail API key). Input file: CSV with columns first_name, last_name, company_domain.

I have a CSV at [absolute/path/to/leads.csv] with the columns: first_name, last_name, company_domain. For each row in the CSV, use the Findymail MCP to find a verified email. For each lead, return: - first_name - last_name - company_domain - verified_email (or "not found") - confidence_score (0-100, from Findymail) - verification_status (verified / risky / invalid / not_found) Process the entire file. Skip rows where company_domain is missing. Save the enriched output as a new CSV at [same/path/leads_enriched.csv] preserving the original column order and appending the three new columns. After processing, give me a one-line summary: total rows, verified count, hit rate %, and total Findymail credits used.

Pro tipDrop every "risky" verification before upload. The 10–15% list shrink will save you 30 days of warmup damage if even one of those bounces.

Section 04Lead Scoring

OperationScore and tier a raw lead list against your ICP — Hot / Warm / Cold — with reasoning per lead.

Most operators upload everything Apollo gave them and let Instantly figure it out. The reply-rate ceiling on that approach is 1%. Pre-scoring against a written ICP and removing the bottom tier before launch is the single highest-leverage thing you can do to a campaign — and Claude can do it in one pass over a CSV with no MCP.

Prerequisite setupNo MCP required. Have your CSV path and a written ICP definition ready (3–8 firmographic + role criteria, plus 1–2 disqualifiers).

I have a raw lead list at [absolute/path/to/leads.csv]. The columns are: first_name, last_name, title, company, company_domain, employee_count, industry, [+ any others present]. My ICP is: - Company size: [e.g., 50-500 employees] - Industry: [e.g., B2B SaaS, agencies, professional services] - Role: [e.g., Head of Sales, VP Sales, CRO, Founder if <50 ppl] - Geography: [e.g., US/Canada only] - Hard disqualifiers: [e.g., consulting firms, recruiting agencies, anyone with "freelance" or "consultant" in title] Score every row in the file as HOT, WARM, or COLD using these rules: - HOT = matches ALL ICP criteria, role is decision-maker, no disqualifier present - WARM = matches MOST criteria but has one soft mismatch (e.g., role is influencer not decision-maker, or company size slightly off) - COLD = misses two or more criteria, OR triggers any disqualifier For each row append two columns: tier (HOT/WARM/COLD) and reason (one sentence — what specifically tiered this lead). Save the output as [same/path/leads_scored.csv]. After processing, give me a count breakdown by tier and recommend whether to run the campaign on HOT only, HOT+WARM, or all three.

Pro tipNever run a cold campaign on COLD-tier leads to "fill volume." A 500-lead HOT-only campaign beats a 2,000-lead mixed campaign on every metric that matters.

Section 05CRM Pipeline Updates

OperationUpdate lead statuses in Attio after a campaign reply window closes — no manual entry, no Ops hire.

The work no one wants to do — moving 200 leads from "Contacted" to "No Reply" or "Replied — Not Interested" — is exactly where pipelines rot. Attio's MCP lets Claude run a weekly sync: pull the campaign's reply data, match against Attio records, update stages in bulk, and log a note on each record so your team has an audit trail.

Prerequisite setupAttio MCP installed and authenticated. Know your workspace's lead/people object name and the stage field (often a status attribute). Have campaign reply data exported (or pull it via Instantly MCP in the same session).

Run the weekly Attio pipeline hygiene update. Step 1: Pull the list of leads currently in stage "Contacted" from Attio (use list-records-in-list or search-records with the appropriate filter). Step 2: For each lead in that list, check their last_contacted date. For any lead contacted more than 7 days ago with no reply logged in Attio: - If they replied positively (check my Instantly campaign reply data at [path/to/replies.csv] OR pull via Instantly MCP) → update Attio stage to "Replied — Interested" and add a note with the reply snippet. - If they replied negatively → update stage to "Replied — Not Interested" and add a note. - If they did not reply → update stage to "No Reply — Cold" and add a note "auto-moved after 7-day reply window." Step 3: For each Attio record updated, log a comment on the record: "Stage auto-updated [today's date] based on campaign reply window closure." After the run, give me a summary: total records reviewed, count moved to each new stage, any errors encountered.

Pro tipSchedule this as a weekly Friday cron — your Monday pipeline review becomes a 5-minute meeting instead of a 90-minute Ops scramble.

Section 06Automated Follow-Up Sequencing

OperationIdentify leads stalled in your pipeline and queue contextual follow-up drafts automatically.

Most "no replies" aren't rejections — they're forgotten. A lead stuck in Contacted for 5+ days with no reply is the highest-EV cohort in your CRM, and the message that wins them back almost never lives in your default Instantly sequence. Use Claude to surface them and draft a personalized bump using their original Attio context.

Prerequisite setupBoth Attio MCP and Instantly MCP installed. Know which Instantly campaign or sequence is your "follow-up bump" container.

Run the stalled-lead follow-up generator. Step 1: Use the Attio MCP to pull all people records where stage = "Contacted" AND last_contacted_date is between 5 and 14 days ago AND no reply has been logged. Step 2: For each stalled lead, pull the original outreach context from their Attio record — note history, company, title, and any custom fields about why they fit ICP. Step 3: For each lead, draft a follow-up email that: - Does NOT start with "just bumping this" or "circling back" - References one specific detail from their Attio record (their company, their role, a note on their account) - Is 3-4 sentences max - Ends with a soft CTA (a question, not a meeting ask) Step 4: Output the drafts as a CSV with columns: email, first_name, company, follow_up_subject, follow_up_body. Save to [path/to/followup_drafts.csv]. Step 5: Use the Instantly MCP to add these leads to the campaign named "[your follow-up campaign name]" with the drafted body inserted as a custom variable. Do NOT activate the campaign — leave it paused so I can review before launching. Summary at end: count of stalled leads found, count of drafts written, count successfully added to Instantly.

Pro tipFollow-up bumps that reference one detail from the original context out-perform generic "still interested?" bumps by 3-4x on reply rate. Specificity is the entire game.

Section 07Cold Email Sequence Writing

OperationGenerate 3 distinct angle variants of a cold email sequence — ready to paste into Instantly as A/B/C tests.

Most operators write one sequence and call it done. The operators who win run 3 angles per launch — pain-led, status-led, and proof-led — and let the data pick the survivor by week 2. Claude can produce all three in a single pass if you feed it a tight ICP, the offer, the top 3 objections, and your real proof points.

Prerequisite setupNo MCP required. Have your ICP definition, one-line offer description, top 3 objections (in your prospect's voice, not yours), and 2–3 specific proof points (case study results, metrics, named clients).

Write 3 cold email sequence variants for the following campaign. Each variant should be a 3-step sequence (initial + 2 follow-ups). ICP: [paste ICP definition] Offer: [one sentence — what we do, for whom, with what outcome] Top 3 objections we hear: [list — written in the prospect's voice] Proof points: [list specific outcomes, metrics, or named clients] Build 3 angle variants: VARIANT A — PAIN-LED Lead with the specific operational pain this ICP feels weekly. Use objection #1 as the wedge. Proof point goes in step 2. VARIANT B — STATUS-LED Lead with what their best-in-class peers are already doing. Use a named-client proof point in step 1. Reframe objection #2 as a competitive risk in step 2. VARIANT C — PROOF-LED Lead with a specific outcome metric (e.g., "we took [client] from X to Y in Z weeks"). Reverse-engineer the curiosity. Address objection #3 in step 2. For each variant, output: - Subject line (3-5 words, lowercase, no caps gimmicks) - Step 1 body (under 75 words, no "Hope this finds you well") - Step 2 body (under 60 words, sent 3 days later) - Step 3 body (under 40 words, breakup email, sent 5 days after step 2) - CTA style used (soft question / direct meeting ask / resource offer) Format the final output as a markdown table I can read side-by-side, then below the table give me each variant in plain text formatted for direct paste into Instantly's sequence editor.

Pro tipKill the worst variant by day 7 based on open rate, kill the next one by day 14 based on reply rate. Don't run all three for a full month — you're losing learning velocity.

Section 08Campaign Orchestration

OperationCreate, populate, and launch a full Instantly campaign end-to-end from a single Claude Code session.

This is the payoff. Once your list is enriched (Section 3), scored (Section 4), and your sequence is written (Section 7), the entire campaign build — creating it in Instantly, importing leads, attaching the sequence, setting the schedule, mapping sender accounts — is one prompt away. What used to be a 45-minute manual setup becomes a 90-second Claude Code run.

Prerequisite setupInstantly MCP installed and authenticated to the correct workspace. Have the lead CSV path, the sequence text from Section 7, the campaign name, and the send schedule (timezone + send window) ready.

Build and launch a new Instantly campaign end-to-end. Campaign name: [e.g., "2026-04 — HOT — Mid-market SaaS — Variant A"] Lead CSV path: [absolute path to scored + enriched CSV — HOT tier only] Sequence: [paste the 3-step sequence from Section 7, Variant A] Send schedule: - Timezone: America/New_York - Send days: Tuesday, Wednesday, Thursday - Send window: 9:00 AM – 4:00 PM - Daily limit: 30 emails per sender account - Stop on reply: yes - Open tracking: off - Link tracking: off Sender accounts: use all currently warmed accounts in the workspace (pull the list via Instantly MCP and attach all of them). Steps: 1. Use Instantly MCP to create_campaign with the parameters above. 2. Use create_lead_list and add_leads_to_campaign_or_list_bulk to import the CSV. Map columns: email → email, first_name → first_name, last_name → last_name, company → company. Skip any row missing email. 3. Attach the 3-step sequence to the campaign. 4. Configure schedule and sender accounts. 5. Do NOT activate yet. Pull the campaign back via get_campaign and show me the full configuration as a checklist so I can verify before launch. After I confirm, activate the campaign with activate_campaign and confirm it is live.

Pro tipAlways do the dry-run + checklist confirmation step before activating. Instantly does not have an "undo launch" — and a misconfigured campaign with the wrong sequence on 500 HOT leads is the kind of mistake that costs a quarter.

ClosingRun this stack this week

You now have eight prompts that replace what most agencies pay an Ops hire $80k+ to do. None of these require an engineer, a Zap, or a custom integration. They require an MCP install and a session.

Want this mapped to your offer?

Book a free 30-minute discovery call. We'll map your current outbound to this playbook and show you exactly which two prompts will move your reply rate the fastest — including the LSN sourcing, Findymail enrichment, and Instantly orchestration we run for our own clients.

Book A GTM Strategy Call →

← Back to all guides

The Claude CodeGTM Playbook.