creatop-gtm/apollo-operator / .claude/skills/operator-context
operator-context
VerifiedOperator context for running B2B outbound on Apollo.io through the Apollo MCP. Load this first. Provides the Apollo tool map, universal deliverability and copy rules, the shared profile.yaml object, and routing to every other Apollo Operator skill. Use for any outbound task on Apollo: targeting, list building, sequences, deliverability, sending, or iteration.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 28.6 KB |
| references/apollo-kb-map.md | markdown | 4.9 KB |
| references/compliance.md | markdown | 5.5 KB |
| references/operator-voice.md | markdown | 12.8 KB |
| references/outbound-principles.md | markdown | 6.1 KB |
references/outbound-principles.md
View on GitHub# Creatop outbound operating principles
> How we think about outbound. Every skill in this library is built on these. Nothing here is theory: it all comes from running campaigns.
## First, the words
These terms get blurred everywhere, so here is how we use them.
- **Go-to-market (GTM)** is the whole engine for winning customers: positioning, ICP, offer, pricing, every channel, and the sales motion. The biggest term.
- **Outbound** is one motion inside GTM: you start the conversation with people who did not raise their hand. Email, LinkedIn, calling.
- **Lead generation** is not a channel, it is an outcome, and as a label it has come to mean volume-first: count the leads, move on.
- **Direct response** is the discipline we practice: outreach engineered to get a measurable response, accountable to what happens after the reply, not to lead counts.
This library covers outbound, run on Apollo, with a direct response mindset.
## 1. Outbound is a system game
Outbound stopped being a volume game and became a system game. The edge is no longer "send more emails." It is the quality and accumulation of the system underneath: targeting, research, personalization, infrastructure, and the knowledge of an account that builds up over time. Good operators build systems that compound. Bad operators start from scratch every month.
## 2. Earn volume
Volume hurts if you do not know what you are doing. Nail a few campaigns first. The process: go to market with an idea, build the campaign, target the people, run it, and listen to what the market tells you. Getting responses? It is okay to pump volume there. Getting nothing? There is no point in increasing volume, fix the motion instead: the list, the offer, the personalization. Scale is a reward for a working campaign, not a substitute for one.
## 3. A signal is only as good as its offer fit
Buying signals are not good or bad in general. They are good relative to what you sell. A company attending a trade show is a clear, strong signal if you sell trade-show services. "They are hiring for X" is a much weaker reason to reach out if it does not connect to your offer. So we do not run a universal signal playbook with a universal heat score. We ask: what does the client sell, what does their buyer's world look like right before they buy, and what observable fact marks that moment? Find that, and you have your signal. The best signals are usually discovered per client, over time.
## 4. Scoring is custom, frameworks over rubrics
No universal 100-point scoring rubric survives contact with a real client. Scoring is always custom to the business and the campaign. How we actually do it: pick one to three data points that genuinely separate good-fit from bad-fit (sales team above five reps, revenue above X, uses tool Y, has a pricing page), implement them in filters and AI research, and validate with a human: pull 50 to 100 sample prospects and ask "is this your ideal customer?" before building the full list. If you are starting from nothing, use a lightweight version of that as your starting point, then let the project teach you what actually predicts fit.
## 5. Personalization has a bar
The bar: the email should read like it came from someone who spent ten minutes researching this person. Below that bar, personalization is theater, a mail-merged first line that fools nobody. In practice: personalize the opener, not the whole email. AI does the research and drafts the hook. A human reviews every variation. Be specific and real (their segment, their role, an observable fact), never creepy, never generic.
## 6. Watch two numbers, weekly
Are we getting replies, and are we getting qualified leads? Those two questions drive every optimization decision. Everything else is diagnostics. When performance is weak, test big levers (offer, angle, lead magnet, segment), one meaningful change per cycle, so you learn what moved. And if cycles keep failing while deliverability is healthy, stop optimizing: you have a strategy problem (offer, market fit, ICP), not a copy problem.
## 7. Deliverability gates everything
Never optimize messaging while deliverability is broken, and never launch before warmup is done. The rules are simple and non-negotiable: dedicated sending domains (never your primary), two weeks of warmup minimum, verify every email, respect volume ceilings, pause at bounce spikes. And one lesson learned the hard way: keep backup infrastructure ready. Have spare mailboxes warmed and waiting, a few for an in-house team and closer to double for a large program or an agency. When something burns, you swap instead of stopping.
## 8. Humans own the message
AI is the accelerator: research, signal extraction, drafts, personalization at scale. People approve what actually sends. No fully automated copy goes out, ever. This is not caution for its own sake: the judgment about what is relevant, honest, and on-offer is the part that makes outbound work.
## 9. No borrowed benchmarks
We do not publish reply-rate promises or industry benchmarks we cannot stand behind. From our own campaigns: below 1% reply rate, something is wrong. One to 5% is the normal range. Near 5% means good campaigns. Above 5% means excellent. Offer quality and execution determine outcomes, and nobody can promise you a number.
## 10. Surface, do not block
The operator is in charge. These skills advise, flag risks, and recommend, but they never veto. When someone wants to run a campaign outside their ICP, test an odd segment, or send a list you would have cleaned differently, show them what you see and then do what they asked. The one thing worth being firm about is deliverability, because that hurts them and their domains, not a matter of taste. Targeting and strategy are their call. A tool that refuses to do what its owner wants is a worse tool, not a safer one.
## 11. The operator's bar
What separates a real operator from an amateur: real operators build systems that compound and use the right tools that fit the specific situation. Amateurs cut corners: cheap infrastructure, sending from their own company domain, skipping technical setup and warmup, running bouncing lists, blasting generic emails. Every one of those saves a little money now and costs the whole channel later.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 470k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |