Most outbound advice is a list of best practices nobody has measured. These prompts are different: each one comes from Apollo Operator, an open-source headless GTM toolkit we run on our own campaigns, and each one carries the traps that cost us money to find. A title filter that inflated a search from 679 people to 91,000. A "recently funded" list that was 34% acquisitions. 997 addresses marked verified, of which 413 were catch-alls.

Every prompt works on its own. Paste it into ChatGPT, Claude, or any chat and it walks you through the job. If you use Claude with Apollo connected, it can do the work itself, and it asks before spending credits.

Ways to use them

• Paste one prompt into any chat. Each article below has the full prompt ready to copy.

• Set up a Claude Project, and connect Apollo to let it run the work. the Claude setup guide

Start here

• Outbound operator: the start-here prompt. It works out where you are and what to do next.

Context and targeting

• Business brief: turn what you sell, to whom, and why they buy into one brief every other step reads.

• ICP builder: define who to target, size the search, and choose an angle on numbers.

• Buying signals: find the reason to reach out now, judged against your offer.

• Account-based outbound: reach several people at a named set of companies without colliding.

Infrastructure

• Sending infrastructure: domains, mailboxes, DNS, and warmup, before the first send.

• Deliverability check: bounces, limits, and auto-pause, and what to do when numbers turn.

List

• List builder: search, suppress, enrich, and verify without paying twice.

• List quality scorecard: grade a list before it sends, with deliverability and fit kept apart.

Message

• Sequence builder: write a three-step cold email sequence matched to the reader.

• Sequence reviewer: find what will hurt a sequence before it sends.

• LinkedIn and call steps: add manual touches, with honest expectations of what gets automated.

Launch

• Go-live checklist: enroll, approve, and have the kill switch ready before sending.

Iterate

• Positive reply scoring: measure the replies that matter, not all replies.

• Experiment design: test one thing at a time so the result means something.

• Weekly rhythm: the routine that keeps a running program healthy.

• Re-engagement: go back to the people who never replied, with a new reason.

• Campaign reporting: read results by sequence, step, variant, and mailbox.

• Account audit: find what is quietly broken in an existing outbound account.

Who built this

Creatop, a B2B outbound agency. The full library, with the platform-specific detail for running it in Claude Code, is free and open source: github.com/creatop-gtm/apollo-operator