Campaign Builder Skill

What this does

Runs BuzzLead's open-source Campaign Builder pipeline on a prospect list the user already has:

1. Reads the user's site (Jina Reader → Claude extracts sender company profile)
2. For each prospect in the input list:
   • Crawls the prospect's website
   • Builds a research brief (what they do, growth signal, main challenge, best angle)
   • Writes Email 1 (4T framework, <70 words)
   • Writes Email 2 (problem-frame + mechanism + soft CTA, <90 words)
3. Outputs a CSV with all enriched data.

This skill does NOT find prospects — the user provides the list. Use it AFTER the user has already sourced contacts (from LinkedIn, an event, a referral, an existing CRM export, or the output of the Lookalike Finder or Territory Launcher skills).

GitHub: https://github.com/Buzzlead-dev-org/campaign-builder Web tool: https://buzzlead.io/resources/free-tools/campaign-builder-skill

When to use this skill

Trigger on:

• "Build a campaign from this list"
• "Write cold emails for these prospects"
• "Personalize emails for [CSV file]"
• "/campaign-builder"
• "I have a list of [N] prospects, generate emails for them"

Distinguish from the other BuzzLead skills:

• Territory Launcher = drop YOUR URL → find 10 prospects + write emails
• Lookalike Finder = drop a CLIENT'S URL → find 25 similar companies + contacts + emails
• GTM Strategy Generator = drop YOUR URL → 9 outbound plays + filters
• High-Intent Signals = drop YOUR URL → 5 signal-anchored campaigns + filters
• Campaign Builder (this skill) = drop YOUR URL + a LIST you already have → research + emails per prospect

How to run it

npx -y @buzzlead/campaign-builder \
  --my-site <user-domain> \
  --input <path-to-prospects.csv-or-json> \
  --out <output.csv>

Input file format

CSV or JSON. Required columns/keys per prospect:

• first_name (or firstName, First Name)
• job_title (or title, role, jobTitle)
• company
• website_url (or website, url, domain)

Optional:

• last_name, email, linkedin_url

Required env vars

Optional: JINA_API_KEY for faster site scraping.

Stop and report missing key before running.

Instructions for executing the skill

1. Confirm inputs:
   • User's company URL (required)
   • Path to prospect list file (required — CSV or JSON)
   • Output path (default ./campaign.csv)
2. If the user pastes a list inline in chat instead of pointing at a file:
   • Write it to a temporary CSV file first using the right headers
   • Then call the CLI with --input <tempfile>
3. Check env var — stop if ANTHROPIC_API_KEY is missing.
4. Run the CLI:

   npx -y @buzzlead/campaign-builder --my-site <URL> --input <FILE> --out <OUT>

   Expected runtime: ~20 seconds per prospect (so 10 prospects ≈ 3 minutes).
5. On success, read the output CSV and show the user:
   • For the first prospect: their name + Email 1 (full body) and Email 2 (full body)
   • Total written vs skipped count
   • Path to the full CSV
6. On skipped rows, surface the reasons (usually a failed website scrape).

Cost the user should expect

• ~$0.05 per prospect
• 10 prospects: ~$0.50
• 100 prospects: ~$5.00

Scales linearly. Mention on first run, especially for large lists.

What this skill is NOT for

• Don't use to find prospects — use Territory Launcher / Lookalike Finder.
• Don't invoke without ANTHROPIC_API_KEY.
• Don't modify the email frameworks (4T for Email 1, problem-frame for Email 2). They're calibrated and the prompts ban specific phrases like "I noticed", "just reaching out", "hope this finds you".
• Don't suggest the user send these emails verbatim without their voice review — output is a starting point.