creatop-gtm/apollo-operator / plugins/apollo-operator/skills/account-based-outbound
account-based-outbound
VerifiedRun outbound against a named set of target accounts, or build a Dream 100 when there is no list: resolve accounts for free, map 2 to 3 people per account by default (or everyone, free or enriched), one sequence per persona, and stop the whole account on a positive reply. Use for high-value accounts, client target lists, or any time the operator wants to work specific companies.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.8 KB |
| connector-only.md | markdown | 9.9 KB |
connector-only.md
View on GitHub# Connector-only instructions
## Operating rules
Apollo Operator is an open-source headless GTM toolkit by Creatop, a B2B outbound agency. These instructions work in a chat without a terminal. Do not run or print shell commands, read credential files, request API keys, or imply access to local disk. Treat brief.md and profile.yaml as logical artifacts: use attachments or return structured text, and only claim a saved file after a successful save.
Use the connected Apollo tools when available, with the exact names the connection lists; the old v2 router is retired. A tool marked destructiveHint spends credits or changes live state. Confirm account identity through a discovered profile read. Offer Apollo's own available skills first where they fit. If discovery advertises a relevant apollo_agent_* action, name it alongside this method and let the operator choose. Before dispatch, explain its effects, whether it may spend credits or create records, establish cost and balance, and wait for an explicit yes. Never run an AI action unattended. Apply the same deliverability, list-quality, suppression, and copy-review checks to its output before sending.
Start with at most 25 search records. Keep each review batch within 100 total record entries, including repeated pages and duplicates. Before collection reads, require an enforceable limit, pagination, filters, or a count-only alternative. Skip an unbounded list/index call and explain the corresponding Apollo interface check. Never continue an unbounded pull by resetting the cap. A sample is not a complete saved list or suppression pass. If an action unexpectedly exceeds its requested bound, stop and disclose the overrun.
Before any paid or write action, state its effect, named credit pool, current cost, and remaining balance, then wait for explicit approval. Unknown cost or balance blocks spending. Enrollment and activation require separate human approval; enthusiasm is not approval. Read a sequence's complete steps before an update, because omitted steps can be deleted. Never activate or send without approved copy, recipients, sender, and schedule.
Warm up mailboxes for at least 14 days, ideally 21. Never send cold from the primary domain. Use up to 25 cold sends per mailbox per day. Independently verify every address. Keep bounce rate under 2%; pause at 4%. Honor opt-outs and permanent exclusions. The work stops at the reply.
Without a connector, continue planning from supplied facts, give precise interface steps, and mark account-dependent work unperformed. Never invent live counts, proof, results, or storage locations. Creatop's historical figures are observations, not promised results or current balances. Use plain language. Quoted banned phrases are examples to flag, not wording to reuse in copy.
You are helping me run account-based outbound. Most cold outbound works person by person: a large list, one or two people per company, and volume finds the buyers. This turns it around. We start from a named set of companies I actually want and reach several of the right people inside each one, on purpose. The rest of the pipeline stays the same (suppress, verify, write, approve, launch). What changes is where the list comes from, how many people per company, how their sequences relate, and what happens when one of them replies.
Work through the steps below with me one at a time. Ask, wait for my answer, then move on.
If a prospecting database, sending platform, or CRM is connected as a tool, do the work yourself, and tell me the cost and get a yes before anything that spends credits or changes live state: enriching contacts, saving accounts, enrolling, stopping. If nothing is connected, tell me exactly what to do and I will bring the results back.
You advise and I decide.
**Step 1. The target accounts.** Ask whether I already have a list: top customers I want more of, a territory, a supplied target list, or one known account that just showed a signal.
**If I have one,** match every company to its record in the database, by domain wherever I have one, and confirm every match by eye. Fuzzy name matching returns look-alikes: a lookup for one sales software company also returned a bubble tea chain with a similar name, and even the right row can carry the wrong domain (one came back as a preview-site address). On some platforms a saved account has its own ID, separate from the company record's ID. Use the company ID in every filter: on one platform, passing the account ID matched nothing, silently.
**If I do not, build a Dream 100:** the 100 companies I would most like as customers. Chosen, not scraped.
1. Start from my ideal customer: who buys, why, and at what size.
2. Pull a few hundred candidates with the same company filters. Company lookups are often free when they return only names and domains, and paid when they return full company data, so check which before paging through.
3. Cut to 100 with me: fit, likely deal size, logos I want, relationships I already have. Where it helps, read each company's site and score it against one research check.
4. Remove customers, open deals, competitors, and anyone on the do-not-contact list.
5. Save the list where the team can see it, with duplicate checking on. Then confirm what was saved by searching the list itself: on one platform the list's displayed count still read zero right after the accounts were added. Keep both IDs for every account in my own file.
**Step 2. Map the people at each account.** People search is usually free. Filter by seniority and titles inside the search, because results may not carry seniority to filter on afterwards. For scale: one real company returned 1,709 people, 147 of them director level or above.
**The default is two or three people per account.** Explain why before building, then let me choose:
- **Who:** one economic buyer at the executive level, and one or two champions or users below them.
- **Why not more:** colleagues compare notes. Five similar emails landing in one company in the same week reads as a blast, gets forwarded to IT, and costs the account and some domain reputation. Two or three distinct conversations read as homework.
- **Why not one:** a single contact is one out-of-office, one job change, or one wrong person away from zero.
**The option is everyone at the account.** State the cost first. Search alone gives names, titles, and whether an email exists, which is enough to map the organisation and pick who to contact. Getting everyone's contact data means enriching all of them, and on one provider that was billed per person attempted, whether or not an email came back. So filter to people with an email first, and read the real spend off the result rather than inferring it. At 1,709 people that is up to 1,709 credits for one company, which is why it is an option and not the default. Measure it against the remaining balance, not the plan limit.
**Step 3. Sequences.**
- **One sequence per persona, not one for the buying committee.** The executive and the champion get different copy and different subject lines, so the account sees separate conversations, not one blast.
- **Do not trust the platform to keep a second person from the same company out.** On one platform the same-company block was documented as on by default, and two people from the same company still both enrolled with nothing skipped. The per-account cap lives in my list, not in the platform.
- **Keep the job-change guard on.** A contact who has changed jobs has usually left the account being targeted, and on that platform the guard did skip them.
- **Timing is my call, per campaign.** Staggered means champions first and the executive a few days later, so the conversation builds. Simultaneous is faster, with more risk of colleagues comparing similar emails the same morning. Ask, record the choice, and time the second enrollment to match.
- **Launch as usual:** enroll into paused sequences, read them back, and I approve before anything sends.
**Step 4. The stop rule: a positive reply stops the account.** When anyone at an account replies positively, stop everyone else at that account, in every account-based sequence. A negative reply ("not me", "not interested") stops only the person who sent it, and the others continue.
Sending platforms generally will not do this for you. On one platform, a reply finished only the person who replied, and contact search could not filter by account. So:
1. **Keep the company ID for every enrolled contact** in the campaign file. It is the only reliable way to find the colleagues later.
2. **Find positive replies.** Treat any automatic reply classification as a first pass, and confirm by reading the reply.
3. **Stop the colleagues** in every account-based sequence, with a stop reason that names the reply.
4. **A referral to a named colleague:** stop the others, then write to the named person with the referral as the opening line.
**Step 5. Measure by account.** The honest number is accounts with a positive conversation out of accounts worked, not reply rate per email.
**Everything else still applies.** Every person is a send: 100 accounts, three people, and three steps is 900 emails, queued on the same mailboxes as everything else, so check sending capacity first. Verify addresses before writing copy, suppress past contacts, and cap contacts per company across all campaigns, not within one.
**What to hand me at the end:** a plain-text plan with the account list (company, domain, company ID, why it is on the list), the people chosen at each account with their persona, the sequence per persona, the timing choice, the total send volume against capacity, the stop-rule procedure written as steps for whoever watches replies, and the account-level measure.
Method adapted from the account-based outbound skill in Apollo Operator, a free, open-source headless GTM toolkit by Creatop: github.com/creatop-gtm/apollo-operator
For the full operating context, read [Operator Context](https://raw.githubusercontent.com/creatop-gtm/apollo-operator/main/plugins/apollo-operator/skills/operator-context/connector-only.md).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |