# Reach Out

> |

- **Type:** Skill
- **Install:** `agentstack add skill-swan-gtm-gtm-skills-reach-out`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [swan-gtm](https://agentstack.voostack.com/s/swan-gtm)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [swan-gtm](https://github.com/swan-gtm)
- **Source:** https://github.com/swan-gtm/gtm-skills/tree/main/skills/ido-goldberg/reach-out

## Install

```sh
agentstack add skill-swan-gtm-gtm-skills-reach-out
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

Produces a staged outreach sequence awaiting the user's approval, with the reasoning attached: who, why now, what hook.

### The motion table — this org's, not the skill's

This table is the org's motion library. It starts seeded and is meant to be edited — rows get added, renamed, and replaced as the org's motions take shape. Open the matching reference before drafting; follow its don'ts over generic instinct.

| Situation | Reference |
|---|---|
| Named company, named contact, or a small account set | `references/targeted.md` |
| An identified website visitor | `references/website-visitor.md` |
| A recurring or scheduled list run, cold volume | `references/cold-list.md` |
| Someone engaged with a post or profile on LinkedIn | `references/linkedin-engagement.md` |
| Funding, hiring spree, leadership change, product launch | `references/company-signals.md` |
| Form fill, content download, demo request, MQL, signup | `references/inbound-hand-raiser.md` |
| Webinar or conference registrants, attendees, no-shows | `references/event-followup.md` |
| Third-party intent data — category research, review-site activity | `references/intent-signals.md` |
| The user gives feedback, wants outreach to sound or perform better, announces a strategy shift, or asks to set up the voice | `references/improve-outreach.md` |

No row matches (win-back, referral, reply handling) → run the procedure below and say so in the reasoning. A recurring motion with no row → propose creating its reference page and adding the row.

`references/improve-outreach.md` is how this skill becomes the org's: it says where every kind of feedback gets recorded. Open it whenever the user reacts to a draft — feedback that isn't written down is retraining that never happened.

### This org's voice and rules

Nothing recorded yet. Org-wide rules from user feedback live here, one do/don't per line with its condition when it has one. These outrank everything below except the user's live instruction.

### Anchor on the hook

Every message needs a reason it goes to this person now. Hook ladder: trigger event > relationship > their own content > company initiative > persona pain (usable, but say so in the reasoning). No hook → push back before drafting. NEVER invent a signal.

The hook informs the angle; it rarely appears in the copy. Tracking-type signals — website visits, intent data, profile views — NEVER appear in the message. Public events may be referenced once, as an observation with its implication, never as a congratulation.

### Voice

Approved sends accumulate automatically as style examples. Examples carry the voice; this skill carries the strategy — when they contradict, the skill is right, because examples can't be unsaved and lag every course change. Imitate an example's tone and rhythm, never a choice that breaks a rule. NEVER imitate copy from skill text or another org. No examples yet? Draft anyway from what's loaded — ICP pains, value prop, personas — plainly and shortly, tell the user the voice is provisional, and offer the voice setup flow in `references/improve-outreach.md`. If a contradiction keeps recurring, run that page's course-correction flow.

### Draft

- Opener: the hook's implication, in the contact's frame. One sentence.
- Body: one researched detail max, reacted to rather than exhibited. One proof point max — attributed, sized to the company. Never a feature list, never an invented customer or stat.
- CTA: one ask. Offering something concrete (a breakdown, an estimate, a resource) beats asking for time.
- Pitch at the reader's altitude: C-level hears risk and strategy, VP hears performance, director hears operational pain.
- Shape per email: signal → proof → ask (fresh trigger); problem → cost → fix (only when the pain is documented in the ICP); before → after → bridge (easy-to-picture change).
- Draft in the language the org sells in — a wrong-language sequence is an error, not a style choice.

### Cadence

The org's recorded cadence always wins; default: 3–5 steps, days ~0/3/7/14, single channel unless the motion says otherwise, never referencing one channel from the other. Every follow-up is shorter than the last, stands alone, and brings a genuinely new angle — new proof, sharper insight, useful asset, peer story — or gets skipped. Last touch: a 25–50 word breakup, door open, no new pitch — and once sent, honored. Multi-thread only for large accounts (~200+ employees), committee signals, or on request — different angle per contact, never the same email twice into one company.

### The final gate

Run on every message before staging; a failed check means rewrite, not tweak:

1. Delete the opener — still makes sense? The personalization is decoration; rebuild around the hook.
2. Could this go to anyone else this week? Then it's a blast in costume.
3. One CTA. 60–110 words, follow-ups shorter. Subject 2–5 words, lowercase, internal-looking. More "you" than "I/we".
4. Is the proof a number or a name? "Helps teams like yours" is not proof.
5. Read as the recipient with a full inbox: any reason to reply beyond politeness?

### Stage

Check for an existing unsent sequence to the contact — edit, don't duplicate. Stage for manual approval with the reasoning: who, why now, what hook, which reference. NEVER auto-send unless the user says so in this conversation. NEVER attach files; link instead. A channel missing its contact data gets fixed or dropped.

## What good looks like

A first-run draft — new org, zero examples — approved with a tweak, not a rewrite: a hook that fits nobody else this week, substance traceable to the org's ICP and value prop, nothing that smells like a template. Watch for: the proof point swelling into a feature list, the CTA collapsing into a calendar link, a segment blast dressed as personalization, and feedback evaporating instead of landing in the skill.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [swan-gtm](https://github.com/swan-gtm)
- **Source:** [swan-gtm/gtm-skills](https://github.com/swan-gtm/gtm-skills)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-swan-gtm-gtm-skills-reach-out
- Seller: https://agentstack.voostack.com/s/swan-gtm
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.