Research prospects, write openers that land

SDR/BDR / GTM engineer / content-led founder / demand-gen marketer · Assemble in ~15 min; first run in <2 min once installed.

GTMEdit flowon X● golden 87

step 1ready

Define ICP + target list

outbound-email-strategy8 alts

step 2connector

Pull lead/contact data

⚡ Apollo.io

step 3ready

Deep per-prospect research → brief

customer-discovery12 alts

step 4partial

Verify research signals against official sources

outreach-fanout-research31 alts

step 5ready

Write bespoke personalized opener

cold-email-outreach6 alts

step 6ready

Build follow-up cadence + deliverability/compliance

cold-email-outreach6 alts

step 7connector

Send + log to CRM/inbox

⚡ CRM (HubSpot / Salesforce / Sheet)

Did this recipe work for you?

What you get

▸Research brief per prospect (recent activity, angle, context)
▸Personalized opener/email/DM per prospect, ready to send

tools: your lead list (or Apollo from $49/mo)

tokens: ~$2-5 per 50 prospects

≈ $0-60/mo

What stays yours

· Final read of each opener before sending (5 sec each)
· Provide the lead list or connect a source

✓ Last verified 2026-06-10claude-sonnet-4-6 (subagent, test-mode fixtures)

✓ step 1 · Define ICP + target list

✓ step 2 · Pull lead/contact data

✓ step 3 · Deep per-prospect research → brief

✓ step 4 · Verify research signals against official sources

✓ step 5 · Write bespoke personalized opener

✓ step 6 · Build follow-up cadence + deliverability/compliance

✓ step 7 · Send + log to CRM/inbox

view sample outputs from this run

step 1 · Define ICP + target list

# Lead List — Step 1 Output
<!-- Produced by: outbound-email-strategy skill · Step 1: Define ICP + target list -->
<!-- Input: fixtures/prospects.csv (icp-criteria.md and target-list.md not present — ICP derived from fixture signals) -->

---

## ICP Definition

| Element | Definition |
|---------|-----------|
| **Title** | Founder, CEO, Head of Growth, or Managing Partner (decision-maker with budget authority) |
| **Company type** | Small digital agency or growth shop, 5–25 people, B2B-focused clients |
| **Vertical** | Web design / SEO / performance marketing agencies serving SMB clients |
|…

step 3 · Deep per-prospect research → brief

[SIMULATED CONNECTOR — test mode]
<!-- Step 3: Deep prospect research — no live WebSearch called (TEST-MODE). All signals derived from contact-data.md and prospects.csv fixture. Per skill instructions: "If the file does not exist [or cannot be fully executed], perform the step from its description with professional defaults." Skill file truncated at Step 4b; Steps 1–4a applied in full. -->

# Research Brief — Step 3 Output

Deep per-prospect discovery briefs for all 5 leads. Ordered by outreach priority.

---

## Brief 1 — Sofia Marino · Vetro Digital

### Company snapshot

| Field | Detail |
…

step 4 · Verify research signals against official sources

[SIMULATED CONNECTOR — test mode]
<!-- Step 4: Adversarial verify pass on research-brief.md signals. No live web fetch available (TEST-MODE). Verdicts apply the skill's HARD RULES: a signal is confirmed only if it was literally seen on an official source fetched this run; otherwise it is downgraded/rejected per the adversarial-verify schema. Contacts that were pattern-filled rather than fetched are cleared and flagged lookup_needed. -->

# Verified Signals — Step 4 Output

Adversarial verify pass across 5 prospects. Each contact and each trigger signal is evaluated against the skill's trusted-…

Before you start

needClaude Code installed
needYour lead list CSV (name, title, company) — test mode uses the bundled 5-prospect fixture

Test mode needs none of the optional items — fixtures are bundled.

The problem

When I prep outreach, I want an agent to research each prospect (recent posts, funding, role changes, tech stack) and draft a genuinely personalized opener/email/DM that reads bespoke and doesn't scream AI, so I hit volume without 30+ minutes per prospect or generic spam that gets ignored.

Readiness

The core of this job — ICP framing, deep live per-prospect research into a brief, and writing a bespoke non-AI opener plus follow-up cadence — assembles cleanly from the library (3 ready steps, one verify step partial). What you must bring is the live lead data on the front end (Apollo/Clay) and the send+log+memory tooling on the back end (Instantly/Smartlead + a CRM). The library writes and researches; it doesn't source contacts or send.

Cybernetic layers

Memorybyo

No candidate remembers who was already contacted, prior replies, or which angle was used. Need an external state store — a CRM, a Notion/Airtable tracker, or a sequencer's contact DB — to dedupe and avoid re-pitching across runs.

Background / schedulingbyo

Outbound is inherently scheduled (warm-up ramps, day 0/3/7/14 follow-ups, daily send caps). None of the skills self-schedule; wire to a sequencer's cadence engine or a cron-driven harness run that processes the next batch.

Harnessready

Claude Code runs the skills end-to-end: customer-discovery (#762) fans out research via WebSearch, then cold-email-outreach (#233) drafts opener+sequence. Invoke per-batch (one prospect or a list) from CLI; pipe the brief into the writer step.

Click any step to see what fills it — and the ranked alternatives you can swap in.