Cold Email Ops — Communication skill for Claude Code

cold-email-ops

An agent skill that plans and audits a cold-email program at volume, for agencies running several client accounts or one sender whose arithmetic has to hold.

It answers the capacity question before the first send, then keeps answering it while the program runs: inboxes and domains from a monthly target, burn and replacement cadence, per-domain reply floors, shadow-ban tests, automation error budgets, and per-client reporting.

Install

npx skills add arm-pz/cold-email-ops --skill cold-email-ops

For Qoder, copy the folder into the user skills directory:

cp -r cold-email-ops ~/.qoder/skills/

Ask it things like

• "How many inboxes and domains do I need for 500,000 sends a month?"
• "Reply rates fell on two domains over the last two weeks. Is it the copy or the infrastructure?"
• "Review this AI outbound stack for failure points before we scale."
• "Give me the per-client report for three accounts."

What it produces

Mode	Output
Plan	Capacity, gate design, monitoring spec, and the role/automation map
Audit	Area verdicts (PASS / WATCH / FAIL / INCONCLUSIVE) with the evidence behind each
Stack review	A step map with owner, check, accuracy estimate, and blast radius

The arithmetic ships as a deterministic script, so a plan is not mental math:

python scripts/cold_email_math.py capacity --monthly 100000
python scripts/cold_email_math.py errors 0.8 0.8 0.8 0.8
python scripts/cold_email_math.py ptl --contacts 20000 --leads 53

Defaults are 10 cold sends per inbox per day, 3 inboxes per domain, and 22 sending days a month. At those defaults, 100k/month needs about 455 inboxes and 152 domains; 500k needs 2,273 and 758; 1M needs 4,546 and 1,516.

Operating rules it applies

• Inboxes burn in 6 to 9 months even with clean SPF, DKIM, and DMARC. Replace weekly, or the same day when one fails.
• A domain floor of 1% reply rate. Two consecutive weeks of falling replies puts a domain on watc