Why Claude Skill to run outbound sales campaigns
Most outbound today looks like this:
Use one tool to build lists, another to find emails, another to write sequences, another to send… and somehow you’re the one stitching it all together.
Exporting CSVs. Cleaning data. Copy-pasting personalization. It works — but barely.
Most teams start with a list of names filtered by title and company size. No signal. No urgency. Just hoping someone replies.
What if outbound started with intent — companies that just raised, job posts that match your product, or leaders actively talking about the problem you solve?
That’s exactly what the outbound-prospecting-engine skill does.
It takes you from signal → qualified leads → personalized outreach → live campaign — all in one flow.
Here's how to use this with Claude
You can install this Skill on your computer to use it with Claude Code:
npx goose-skills install outbound-prospecting-engine --claude
#Installs to: ~/.claude/skills/outbound-prospecting-engine/
Alternatively, you can use it directly with Goose using Skills Library
Next, provide the following to your Claude Code:
- Your ICP — target titles, company size, industry, any exclusion criteria
- Your signal keywords — what terms indicate buying intent for your product (technologies, problem descriptions, competitor names)
- Your company context — what you sell, your value prop, your proof points
- Approved messaging or let Claude Code generate sequences based on your context
- Your Smartlead credentials — for campaign launch (or export leads if you use a different sender)
Here's how it works:
1. Detects intent signals across multiple sources
- This is where the pipeline starts — not with a list, but with signals.
- Claude Code runs signal detection across the sources that match your ICP
- You pick which sources to activate. Each one runs in parallel and produces a raw signal list — companies paired with the specific signal that flagged them.
- A company that just raised a Series B AND is hiring for the role your product replaces? That's a multi-signal lead. It surfaces at the top.
2. Qualifies and scores every company
- Raw signals aren't leads yet. Claude Code qualifies each company against your ICP criteria — size, industry, geography, tech stack, exclusion filters.
- Companies that pass get scored. Multi-signal leads (job posting + funding, or LinkedIn activity + conference attendance) score highest. Single social mentions score lowest.
- What you see: a ranked list of qualified companies, each with the signal context that flagged them. Not 1,000 names from a database — the 50-100 companies showing real buying intent right now.
3. Finds the decision-maker contacts
- For each qualified company, Claude Code finds the specific people you need to reach.
- It searches by your target titles from the ICP — not just anyone at the company, but the person most likely to buy, evaluate, or champion your product.
- Pulls email addresses and LinkedIn URLs. If multiple contacts are found at the same company, it flags the multi-threading opportunity.
4. Deduplicates against your contact cache
- Before any outreach happens, every contact gets checked against your existing database.
- Already in a campaign? Skip. Already replied negatively? Skip. Already a customer? Definitely skip.
- New leads get added to the cache. This prevents the embarrassing double-touch — and it means the engine gets smarter over time as the cache grows.
5. Writes personalized sequences using the signal
- This is where signal-based outbound separates from list-based outbound.
- For each lead, Claude Code generates a personalized email sequence using two things: the signal that surfaced them (the "why now") and their company context (the "why them").
- A lead flagged by a hiring signal gets: "I noticed you're hiring for [role] — before your team builds this in-house, here's what [product] does out of the box."
- A lead flagged by a funding signal gets: "Congrats on the Series B. Most teams at your stage start evaluating [category] — here's how [customer] handled it."
- A lead flagged by a LinkedIn post gets: "I saw your post about [problem]. We built [product] specifically for that — here's a quick look."
- The signal IS the personalization. No mail-merge fakery. Real context that explains why you're reaching out right now.
6. Launches the campaign in SmartLead
- Claude Code sets up the full campaign: creates it in Smartlead, uploads the lead list, configures the email sequence (2-3 touches per lead), allocates sending mailboxes, and sets the sending schedule.
- You review the sequences and lead list before anything sends. The skill pauses for your approval at this step.
- Once approved, the campaign goes live. Leads start receiving personalized, signal-based outreach.
The Ongoing Cadence
This isn't a one-shot pipeline. The real value is running it repeatedly:
- Weekly: Re-run signal detection. New job postings, new funding rounds, new LinkedIn activity — new leads enter the pipeline automatically.
- Bi-weekly: Review campaign metrics. Which signals produce the best reply rates? Which sequences are working? Claude Code adjusts.
- Monthly: Review overall pipeline contribution. Which signal sources are worth expanding? Which should you drop?
Every week, the engine gets sharper — the contact cache grows (fewer duplicates), the messaging improves (based on what's working), and the signal sources get refined.
Automate it with Goose Automations
Running the engine manually once a week is already a step change. Running it on autopilot is where it becomes an unfair advantage.
Set up Goose Automations to execute the full pipeline on a weekly schedule. Every Monday, Goose detects new signals, qualifies companies, finds contacts, deduplicates, generates sequences, and stages the campaign for your review. You wake up to a ready-to-send pipeline — not a to-do list.
The human stays in the loop for approval. Goose handles everything else.
Build your outbound engine
The outbound prospecting engine is free and open-source in the Goose Skills Library.
Install it at skills.gooseworks.ai, run it before your next sales call, and walk in knowing more than the prospect expects.
Runs in Claude Code, Codex, Goose or any AI agent of your choice.