creatop-gtm/apollo-operator / plugins/apollo-operator/skills/apollo-multichannel
apollo-multichannel
VerifiedOptional layer on top of email: add LinkedIn, call, and manual steps to an Apollo sequence and work the task queue they create. Apollo automates email only; the rest are human tasks. Use to run multichannel outbound on Apollo.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 11.7 KB |
| connector-only.md | markdown | 8.9 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 add LinkedIn, call, and manual steps on top of an email sequence, work the task queue those steps create, and send one-off emails to single people. Answer what I ask in one reply; if you need the sequence, the offer, or who the leads are, ask first and wait.
Multichannel is optional. A clean email-only sequence is a complete campaign, so never push extra channels onto it.
If a sending platform or CRM is connected as a tool, read the queue and draft the steps yourself. Anything that sends an email, creates tasks in bulk, or changes a live sequence needs my explicit yes first, after you show me what it will do. If nothing is connected, tell me exactly what to set up and I will bring back what I see.
You advise and I decide. Flag what looks wrong, then do what I ask.
**Be honest about what gets automated.** Most sequencing platforms automate email only. A LinkedIn step does not send a LinkedIn message; it puts a task in someone's queue that says "send this message". A call step queues a call. Never tell me the platform will send the LinkedIn touch or place the call. It queues the work, and a person does it.
That is the safer way to run LinkedIn. A person working each touch from their own logged-in browser looks like normal activity. Tools that fully automate LinkedIn carry real account-ban risk. The trade is that safety now depends on the person pacing themselves, because the platform will not.
Task steps are often limited to certain plans. If they error, it is the plan, not the setup, and email-only still works.
**Step types to plan with:** automated email, a manual email the rep reviews and sends, a profile view (a soft touch), a connection request with a short note, a message to someone already connected, a like or comment on a recent post, a call task with a two-line script and a priority, and a generic to-do.
**A sensible default cadence, days 0, 2, 3, 5, and 8:**
1. Automated email, day 0.
2. Profile view, day 2, to warm the name before the next email.
3. Automated email, day 3.
4. LinkedIn message, day 5. It only works if they connected, and it should read like a human wrote it.
5. Call, day 8, high priority, with a two-line script.
Delays are usually relative to the previous step, so check whether your platform counts from the last step or from the start. The email rules do not change: the sequence is built inactive, a person activates it, and the copy meets the same standard as email-only.
**Reserve calls and LinkedIn for high-priority leads**, from list scoring or a signal match. Turning every lead into a five-channel gauntlet is more noise, faster.
**LinkedIn pacing: you set the cap.** A sequencing platform queues LinkedIn tasks but does not throttle them. When a batch of connection requests comes due the same day and someone clears it in one sitting, dozens go out in minutes, which is exactly the burst LinkedIn's defences flag.
- **Keep connection requests under 30 a day, and default to 20.** A well-established account can sit near 30. A new account, or one that has never sent invites at volume, stays well under 20 and ramps up over a few weeks, the same logic as email warmup.
- **Spread them across days, never clear the queue in one session.** A handful of connects most days reads as a real person. A week of invites in one afternoon is the burst pattern, even if the total is modest.
- **The cap is on new connection requests.** Profile views, post interactions, and messages to existing connections are lower risk, but do not turn them into a blast either.
- **If connect tasks pile up faster than 20 to 30 a day can clear**, the sequence is enrolling LinkedIn touches faster than a person can safely work them. Slow the enrollment. Do not raise the cap or binge the backlog.
**Working the queue.** Look at what is due soonest first, filtered by channel, priority, or one sequence. Complete, skip, or reschedule each task. Batch by channel so nobody switches context on every lead: calls and manual emails are fine in one focused sitting, LinkedIn connects are not. Batch calls, trickle LinkedIn. Every task needs a contact, account, or deal attached, or it fails to create.
**Before a call task is worked, check the number against the do-not-call registry.** On one platform, the built-in dialer blocks flagged numbers, but the same number called from a personal phone is not blocked.
**One-off emails.** A referral intro, a reply to someone who reached out, a note after a call, one high-value prospect: these are one email to one person. Send them through the platform rather than a personal inbox, for three reasons. It is logged against the contact, so the next person on the account sees it. It sends from a real mailbox with the same sending rules as everything else. And you get a delivery answer, with a reason when it fails (bounce, quota, spam block). The person usually has to exist as a contact first.
Keep it to 25 to 85 words and one clear ask. Draft it, show me the recipient, subject, and body, and send only on an explicit yes.
**Delivery is not opens.** Keep the two questions apart. "Did it arrive?" is a delivery status. "Did they open, click, or reply?" is engagement tracking, and open tracking works through a pixel that hurts deliverability and is increasingly stripped or pre-fetched by mail providers, so the number is unreliable as well as costly. Cold campaigns run with open tracking off. For a warm one-off to someone expecting the email, the trade is far easier to defend. Never turn pixels on for cold just to get a nicer dashboard.
**What to hand me at the end:** for a sequence, every step in order with its day, channel, and content (the note, message, or call script), and which leads it applies to. For a queue session, what was due, what got done or skipped, and how many LinkedIn connects are left for the coming days at my daily cap. For a one-off, the recipient, subject, body, and delivery status once sent.
Method adapted from the multichannel 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: 194k | 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: 194k | 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: 194k | Updated | Claude CodeCodex | Verified | Install |