Messaging per role
One email and one DM per role, grounded in your lead magnet.
Works on its own in any assistant like Claude, on your own website and data. No MQLooP account, no extension. The MQLooP loop picks it up later if you want it to run.
The prompt, copied as is
Read this skill and run it for my company: https://mqloop.ai/gtm-resources/ai-skills/mqloop-messaging-per-role/SKILL.md
Then paste it into your assistant. It reads the skill and runs it on your website.
Other ways to use it
Claude reads it and can use the skill in the same conversation. In Claude Code, save the file as ~/.claude/skills/mqloop-messaging-per-role/SKILL.md. In Claude.ai, upload the file under skills.
What you get
- A message per role, not one for everyone
- Grounded in the segment's lead magnet
- A sequence, never a blast
Nothing to install besides the file. It never sends anything; with MQLooP connected, every send still waits for your approval.
The skill, in full
This is exactly what your assistant reads. Inputs, method, output format and guardrails.
# MQLooP Messaging per Role
One sequence per lead, never a blast. This skill writes the first touch and the follow-up for
each role in a segment, so the founder, the VP and the manager each read something written for
their seat.
## When to use
- "Write the outreach for <segment>." / "What do we send the VP vs the founder?"
- After the lead magnet exists; the copy points at it.
## Inputs
1. The segment and roles from `mqloop-audience-segmentation`.
2. The playbook title and one-line offer from `mqloop-lead-magnet`.
3. Optional: the MQLooP connector (`get_linkedin_funnel` shows what already went out and what
got replies).
## Method
1. **Seat map.** For each role write the one-line reason they care: founder = time they do not
have; VP = quota and pipeline; manager = the team complains. Same problem, different seat.
2. **Touch 1, email.** Subject under 30 characters that names the asset ("B2B playbook for
<company>"). Body under 70 words: who you are in one line, what you built for them, the
one-line offer, a plain signature, and the opt-out line: Not for you? Reply "no" and I won't
write again.
3. **Touch 1, LinkedIn DM.** Under 150 characters including the link, sent only after a
connection is accepted. The offer only, no pitch.
4. **Touch 2, email follow-up.** Three days later, same thread, a question only,
under 60 words, no link: a credibility line, then "Would you be open to an experiment?" with
the free offer stated once. Never a second asset.
5. **Checks.** One ask per touch. No em dashes. No invented numbers; proof words only from the
company's own results. Pain is never the opener; the solution is.
## Output format
Per role: Subject, Email touch 1, LinkedIn DM, Email touch 2. Then a "Do not say" list for the
segment (competitor phrases, overclaims, words the buyer distrusts).
## Guardrails
- This skill writes copy. It never sends, schedules, or sets a pace; the MQLooP loop does that
inside LinkedIn's limits, and every send waits for the owner's approval.
- Facts about the recipient come from the segment and the public company profile, never from
scraping a person's private details.- 1
Drop in a skill
Download it, or paste its link into any assistant like Claude. No account, no extension.
- 2
Approve or edit the plan
The skill reads your website and drafts your segments and roles.
- 3
Add the extension when it is time to send
One minute in Chrome, on your own LinkedIn. Every send waits for your approval.
Claude Code
~/.claude/skills/<skill name>/SKILL.md
Download a skill, save it there, start a new session. Any other assistant: paste the file in.
Live data, the MQLooP connector
claude mcp add --transport http mqloop https://platform.mqloop.ai/api/mcp
Your token comes from the Account tab on the platform; Claude then runs your loop with MQLooP: segments, leads, approvals.
No Claude? Run it on the web: your work email is all it takes.
