From gtm-skills
Drafts one cold outreach message (email, LinkedIn note, or follow-up) from a signal and persona, with variants and a self-check lint. Reads company voice and proof from a context file.
How this skill is triggered — by the user, by Claude, or both
Slash command
/gtm-skills:atomic-messageThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Draft one outreach message at a time: a cold email, a LinkedIn note, or a follow-up. Generate a primary draft plus 1-2 variants, then lint against the craft rules before emitting.
Draft one outreach message at a time: a cold email, a LinkedIn note, or a follow-up. Generate a primary draft plus 1-2 variants, then lint against the craft rules before emitting.
This skill writes ONE message. It does NOT design multi-touch sequences, cadences, or step arcs, and it does not decide send timing. That is email-prompt-building's job. When a sequence needs each touch written, that skill calls this one per touch.
| Input | Required | Source |
|---|---|---|
| Signal | yes | the reason to reach out now: job change, hiring JD, event, post engagement, funding, or a specific observation. From upstream enrichment, the user, or research. |
| Persona | yes | the recipient's role cluster (e.g. founder, revops, sales leadership, IC). Shapes which pain and proof to lead with. |
| Channel | yes | one of: cold email, linkedin, follow-up. Sets length, format, casing, and CTA style. |
| Recipient context | no | name, company, role facts. Used to make the message relevant, never recited back. |
| Company context file | yes | voice, sender identity, what-we-do, positioning, proof library, CTA options, banned words. Read at runtime. |
Read the company context file's Voice, What We Do, Proof, CTA, and banned-words sections. The skill supplies the craft (how to make an atomic message good); the context file supplies the company's voice and facts. If the context file's Voice section defines a per-channel spec, it overrides the generic channel defaults in references/craft.md.
From the signal + persona, choose exactly one of each:
One signal, one framing, one proof, one ask. If you have three angles, that is three candidate messages, not one crammed message.
Follow the channel's structure and length in references/craft.md, applying the context file's voice. Produce:
Run every draft through the self-check in references/craft.md before emitting. Read each line aloud in your head first. Fix anything it flags.
Output the primary draft and variants as plain text (no markdown blockquotes). For cold email, include subject-line options. Label what each variant changes so the user can choose.
See references/craft.md for the universal copy mechanics, the per-channel structure (cold email / linkedin / follow-up), and the self-check lint.
npx claudepluginhub extruct-ai/gtm-skills --plugin gtm-skillsDrafts personalized cold outreach emails, DMs, and follow-ups using company config, prospect research, Linear deals, and meeting notes.
Activate for: outreach, write email, LinkedIn message, cold email, cold outreach, personalised email, personalise, draft message, first touch, intro email, introduction email, ghostwrite, write for me, email draft, DM, direct message, InMail, connection request, WhatsApp message, outreach message, sales message, prospecting email. NOT for: multi-touch sequences (use sequence), post-call follow-ups (use follow-up), marketing content creation (use content-creation), ad copy or subject lines (use copywriting).
Generates personalized cold outreach messages and email sequences for LinkedIn DMs and follow-ups to book calls and get replies.