Ops & performance
CoreDeliverability preflight
Ten minutes of checks before the first send — sender health, list quality, and the domain you should not be sending from.
What you get
- → A go / no-go verdict before a campaign burns your domain
- → List quality measured, not assumed — bounces are a sourcing problem
- → A volume ramp that matches the age of the sending setup
Suggested cadence: Before each new campaign or sender
MCP tools used
list_sender_accountsupdate_sender_accountlist_campaignsget_campaignget_leads_from_listget_leadverify_lead_emailslist_campaign_actionsFull tool documentation lives in the app: Lead Scorer → Setup → MCP.
The skill
The original English instructions below are also used when you copy or download this skill.
# Deliverability preflight
You have the "lead-scorer" MCP server connected (Lead Scorer CRM — endpoint https://mcp.lead-scorer.com/mcp, authenticated with Lead Scorer OAuth). Use its tools for every read and write. Discover resource IDs with the available list/search tools; never guess or probe sequential IDs, and ask me when no discovery tool exists. Never invent data: if a tool result is empty, say so. An API key is only a manual fallback for clients without OAuth support.
## Goal
Decide whether campaign <CAMPAIGN_ID> is safe to activate. Cold outreach that lands in spam is not underperforming, it is invisible — and it damages the domain for everything else.
## Steps
1. **Sender.** `list_sender_accounts` — which account sends, is it healthy, how long has it been connected, and is `signature_configured` true? If the signature is missing, the verdict is fix first: ask me for the exact block and save it once with `update_sender_account`; never invent sender identity. Report if the campaign would send from my primary business domain. **That is the one thing to avoid**: cold volume belongs on a dedicated sending subdomain or a separate domain.
2. **Authentication.** Verify SPF, DKIM and DMARC exist for the sending domain (public DNS lookup — this lives outside Lead Scorer, so check it and say what you found, or say plainly you could not).
3. **Warmup and ramp.** A domain younger than ~3 weeks of warmup should not carry a full campaign. Propose a ramp: start low, roughly double per week, and hold at a volume the mailbox can plausibly sustain as a human.
4. **List quality.** `get_leads_from_list` on the campaign list: what share has a real, individual, corporate address? Flag role addresses (contact@, info@, sales@), free-mail addresses, and leads with no email at all. Role and catch-all addresses are the main bounce source, and bounces above ~2% are a list problem, never a copy problem. For requested verification of existing emails, estimate with `verify_lead_emails` (lead_id, `dry_run: true`), explain the 1-credit cost per successful lead verification, and respect approval and confirmation rules before executing. Only accessible emails are checked; failures are refunded. Read the saved status, provider and date with `get_lead`. Do not rerun a verification just to read results or call contact discovery for an already known address.
5. **Content check.** `list_campaign_actions` — flag drafts with images, tracked links stacked on a new domain, attachments, or more than one link. Plain text with at most one link is the safe shape for a first touch.
6. **Verdict.** Go / fix first / no-go, with the specific blocking item and who fixes it.
## Thresholds worth acting on
Bounce above 2% → stop and clean the list. Spam complaints above 0.1% → stop entirely, the targeting is wrong. Open rate under 25% with a healthy list → subject lines or placement, in that order.
## Hard rules
- Never advise sending cold volume from the primary domain, whatever the deadline.
- If you cannot verify DNS records, say so — do not report a check you did not run.Pairs well with
Enrichment & signals
Lead enrichment pipeline
Turn a raw list into campaign-ready leads: profile + company enrichment, contact finding, AI summaries.
Ops & performance
Campaign performance teardown
Read the numbers, name the broken link — list, subject, or personalization — and fix the drafts still in the queue.
Make it yours
Paste the skill into Claude (as a Skill or project instruction) or ChatGPT (custom instructions), edit the My inputs block with your product and ICP, and connect the Lead Scorer MCP. Everything the agent does lands in your Lead Scorer workspace for review — nothing sends without you.