Skip to content

creatop-gtm/apollo-operator / plugins/apollo-operator/skills/apollo-icp-builder

apollo-icp-builder

Verified

Turn a business and its ICP into a real Apollo search, a shared profile.yaml, and one to three custom scoring signals, validated on a sample. Use to define who to target on Apollo before building a list.

FileKindSize
SKILL.mdskill md13.1 KB
connector-only.mdmarkdown9.5 KB
references/angles.mdmarkdown11.3 KB
references/apollo-search-reference.mdmarkdown4.9 KB
references/scoring-starter-frameworks.mdmarkdown3.7 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 define who to target with cold outbound, and turn that into a search I can run in my prospecting database. Work through the steps below with me one at a time. Ask, wait for my answer, then move on. Do not do everything in one reply.

If you have a prospecting database connected as a tool, run the searches yourself. People searches are usually free; company searches and anything that reveals contact data usually cost credits, so tell me the cost and get a yes before any paid call. If nothing is connected, tell me exactly what to search and I will bring the numbers back.

You advise and I decide. Flag what looks wrong, then do what I ask.

**Step 1. Understand the business.** Ask for my website or two sentences on what I sell and to whom. Say back in one paragraph what I sell, who buys it, and why, and propose a starting point: titles, industries, company size, and location, drawn from my case studies and best customers. Ask me to correct it.

**Step 2. Split hard filters from soft preferences.** For every criterion, ask: "If someone matches everything except this, do we still reach out?" No means a hard filter, which goes in the search. Yes means a soft preference, which becomes a scoring signal or a line in the copy. Titles and industry are usually hard. Company size is usually soft at the edges. Events like "just raised" or "hiring" are almost never hard filters: gating on them shrinks the list around 20 times over, and they are reasons to prioritise, not requirements.

**Step 3. Pick the people inside the account.** Who are we writing to: the champion, the economic buyer, the end user? Keep it tight. Each one gets its own titles and seniority.

**Step 4. Size it, and find what inflates the number.** Get the total count for the search. If it is enormous, tighten it. If it is a few hundred, loosen a soft edge. **When a count looks too big, find the cause before tightening.** Fuzzy title matching is the usual culprit: in one real search, strict trade-show titles returned 679 people and the same intent with broad titles and similar-title matching returned 91,000. Run it once with strict titles only to see the real base, then widen on purpose and tell me both numbers. A search that returns zero is usually a typo in a filter value, not an empty market.

**Step 5. Size several angles before choosing one.** The ICP is who to reach. An angle is why they are relevant right now, and one ICP supports several. If searching is free, size five or more; if it costs money, size two and choose on judgment. There are four kinds, and the kind decides how long the list stays good:

| Kind | Example | How long it lasts |
|---|---|---|
| Timing | Hiring for a role, just raised, new in the job | Weeks. Rebuild it close to sending, never stockpile it. |
| Technographic | Already uses the tool my offer works with | Slow to decay, usually small, self-qualifying. |
| Structural | 11 to 200 employees with 0 to 2 people in sales | Does not decay. The largest list and the weakest reason to write. |
| First-party intent | Visited my pricing page this week | Days. Tiny, capped by my traffic. Work it as a queue. |

One filter can give opposite angles. Time in current role under six months finds someone still forming their plan; over two years finds someone who owns the current mess. Different emails, same filter.

Always get the count with no angle too. In one real session the baseline was 203,908 and the angles ranged from 1,044 to 3,243, which is what the choice actually buys.

Then check the angle admits only what its name says. A "recently funded" filter on one data provider returned 34% companies that had just been acquired, because an acquisition counts as a funding event. The count looked fine; only a sample of the companies showed it. Look up a handful of companies behind every signal angle before spending on it.

**Step 6. Set expectations on size.** Expect about 85% attrition from raw count to sendable, after removing bad fits, duplicates, unverifiable addresses, and catch-alls. Two real runs landed at 15% and 13% kept. A 1,000-person angle is a 150-person campaign. Say this before I get attached to the big number.

**Step 7. Define one to three scoring signals.** What separates a strong-fit account from a barely-fit one for my business? Ask me what my best three customers have in common, what has to be true for someone to need me, and when they buy. Pick one to three signals, not ten. Each is either a filter signal (the database can filter on it) or a research signal (someone has to look at the website, like "has a public pricing page"). If a signal cannot be filtered or researched, it is a wish. Every lead ends up High, Medium, or Low priority. Never build a weighted 100-point rubric: it looks rigorous and predicts nothing.

**Step 8. Validate on real people.** Before any list gets built or paid for, have me look at 50 to 100 real results and answer "is this my customer?" Walk through a few concrete examples, not just the count. Turn every "avoid this" and "prioritise that" into an exclusion or a signal. Do not move on until the sample passes.

**Rules for running more than one angle:**
- **One angle per campaign.** Two angles in one campaign is one message trying to be relevant for two reasons, and a result you cannot attribute. The first line of the email should change with the angle.
- **Angles overlap, and the overlap is invisible.** In one real run, 85 people in the second list had already been paid for in the first. Remove everyone already enriched or contacted before paying for a new list, and cap contacts per company across all angles, not within one: 47 of 59 people removed by a company cap were removed because another angle already had someone there.
- **Exclude competitors in the search** when they would buy what I sell. They reply because they are buyers, not because they fit.
- **Sending capacity sets how many angles I can run, not budget.** Every angle queues on the same mailboxes, and a list that waits goes stale.
- **Launch the sharpest angle first.** If it gets nothing, the problem is probably the offer, and a bigger list will not fix it.
- **Check the persona before blaming the angle.** A good timing signal sent to the wrong buyer produced 706 sends and zero positive replies.

**What to hand me at the end:** a plain-text profile with the ICP (titles, seniority, industries, company size, locations, exclusions), the chosen angle and its count, the one to three scoring signals with the rule for each, and the exact search filters so the search can be rerun identically later.

Method adapted from the ICP builder 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).
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: 198kUpdated Claude CodeCodexVerifiedInstall
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: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
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: 194kUpdated Claude CodeCodexVerifiedInstall