Claude Code Automates Cold Email Lead Gen End-to-End
Use Claude Code's skills to voice-build Prospeo lists of 1,000 leads, Sonnet sub-agents for zero-extra-cost ICP filtering on SaaS firms, and Karpathy's auto-research repo to autonomously optimize campaigns outperforming humans on 10% volume.
Skills Standardize and Accelerate List Building
Claude Code's 'skills'—simple text files loaded into the tool—encode SOPs, eliminating repetitive prompting and enabling team-wide reuse. Download pre-built skills from repositories like coldoutboundskills (includes cold email copy grader trained on 1,000+ campaigns, Dynadot/Zapier inbox setup, Google Maps/Prospeo scrapers, 12M US businesses, SaaS lists, all US zip codes).
Voice-control via WhisperFlow builds lists without manual filters: Prompt for 1,000 US marketing leaders (CMO titles) at 10-100 employee firms with funding in last 180 days; Claude proposes filters (location: US, funding recency: 180 days, verified emails), exports CSV. Setup takes seconds at code.claude.ai (paste terminal command, no coding needed). This cuts list-building from hours to minutes, accessing Prospeo search/enrich endpoints directly.
Trade-off: Skills handle edge cases (e.g., Zapier API vocabulary) but require initial iteration to refine.
Sub-Agents Hack Delivers Free ICP Filtering and Enrichment
Bypass extra API costs by spawning Sonnet sub-agents within Claude Code's $200/month plan: After Prospeo search, chain to enrich emails/company descriptions, then filter for SaaS/software via sub-agent analysis (e.g., 'Prove they are a SaaS company' on descriptions). Outputs include reasoning, yes/no flags per company—no OpenAI calls needed, all within Claude usage.
Process: Load CSV, enrich (emails + descriptions), batch-classify (e.g., batches of leads), review/adjust classifications interactively. Result: Filtered list of valid-email marketing leaders at confirmed SaaS firms, phones masked unless upgraded. This saves tokens on personalization/ICP checks, scaling to agency volumes (8M emails/month).
Outcome: Hands-free data cleanup/enrichment; sub-agents confirm fits before campaigns, reducing bad leads without separate tooling.
Auto-Research Repo Enables Autonomous Campaign Iteration
Fork Andrej Karpathy's auto-research repo into Claude Code for recursive optimization: Provide business context/rules (e.g., no free services, target profiles), point to senders like SmartLead. Loop: Pull recent results (reply rates), compare to baseline, brainstorm experiments (copy tweaks, list filters), relaunch autonomously.
Agency version visualizes offer loops: Analyzes yesterday's sends, adheres to boundaries, updates copy/lists, stores learnings. Deployed on 10% volume for 2 enterprise clients; one now beats human campaigns. Marketing = testing; this runs experiments 24/7 without intervention, learning optimal filters/copy over time.
Impact: Shifts outbound from manual A/B to AI-driven evolution, biggest lead gen advance in years—install once, let it outperform baselines recursively.