Most cold email audits grade the wrong domain.
You do not send campaigns from your brand domain. You send them from separate warmed domains, and those are the ones that decide whether anything lands. This free agent grades those, names the one thing to fix first, and tells you what leaving it there costs you in replies a month. Ten seconds, no signup, runs on your machine.
Two lines in your terminal.
It installs one skill into Claude Code and touches nothing else. Removing it is one
rm -rf. Needs Python 3.8 and dig, which your machine already has.
curl -sL https://fruitfulsaas.com/outbound-teardown.zip -o teardown.zip unzip -q teardown.zip && cd outbound-teardown && ./install.sh
then, in Claude Code: run an outbound teardown on yourdomain.com
What it hands back.
Not a list of errors. A verdict, ranked by what it costs you, with the fix for each one.
$ scan.py acme.com --sending getacme.com --sending tryacme.com
====================================================================
OUTBOUND TEARDOWN ยท acme.com
====================================================================
COLD SENDING INFRASTRUCTURE 5.5 / 10 (2 domains)
BRAND DOMAIN 9.1 / 10 (acme.com)
Critical / warnings 2 / 3
At 2,000 sends a month from these domains, a healthy setup
earns about 30 positive replies. This one is estimated at 9.
ESTIMATED LOSS: 21 positive replies a month.
--------------------------------------------------------------------
SENDING DOMAIN: tryacme.com
--------------------------------------------------------------------
[CRITICAL] Sending domain is 19 days old
Volume from a domain under 30 days old is the clearest spam
pattern there is. Filters read it as burner infrastructure.
FIX: Warm 3 to 4 weeks at rising volume before campaign one.
[CRITICAL] No DKIM key
Google Workspace uses predictable selectors and none resolve.
FIX: Enable DKIM signing at the provider, publish the key.
Two grades, kept apart.
Mixing them is how audits end up confidently wrong. Everything below is read from public DNS and your public website. No credentials, no account.
What decides inbox placement
- Every domain you send campaigns from, graded on its own
- SPF present, single record, under the 10 lookup limit
- DKIM across 20 selectors, TXT and CNAME
- DMARC and whether it actually enforces
- Domain age, straight from the registry
- Four DNS blacklists on the live mail IPs
Spoofing and business mail
- Whether anyone can send mail that looks like you
- Whether cold volume rides on the domain your invoices use
- Lookalike domains that accept mail, reported never assumed
- Graded separately, and never counted into the reply estimate
Whether it could work anyway
- Whether a stranger can name your buyer from the homepage
- Whether there is proof, or only adjectives
- A booking link at the point of highest intent
Do not name a sending domain and it will not invent a grade for one. A confident number about the wrong domain is worse than no number.
Connect your sending tool
Not sure which domains you send from? Hand it a read-only Instantly or Smartlead key and it reads them out of your mailbox list, so nothing is guessed. It also adds bounce rate, reply rate per campaign, sequence length, warmup state and your real safe capacity. GET requests only. It never sends, pauses or edits anything, and the key never leaves your machine.
On your machine, not ours
The report is a file on your disk. It reaches us only if you decide to send it, and the agent asks once. No tracking pixel, no newsletter, no account. MIT licensed, so read every line before you run it.
Then one of two things.
You fix it yourself. Every finding ships with the actual DNS record or the actual setting. That is the whole point of giving it away.
Or you send the report to me. The agent offers once. I read it and reply within one working day with the fix I would do first, whether or not you hire anyone.
If the fix is bigger than a DNS record. That is a Deployment Sprint: domains, data, sequences, the AI coworker, then the keys.