Lead generation has always been a manual grind.
Find the signal, research the company, track down the right person, dig up their email, verify it, log it somewhere.
Every workflow — no matter how clever the source — bottlenecks at the same point: a human in a browser tab doing enrichment by hand.
That’s exactly the kind of work that AI agents can now handle for you.
This post covers five lead generation workflows that an AI agent like Claude Code, Codex or Goose can execute for you end to end.
Each one is built on a different buying signal and produces a qualified lead list with verified contact data.
1. Turn your competitors’ GitHub repos into a lead list
Every person who stars, forks, or opens an issue on a GitHub repo in your category is telling you they care about the problem you solve. Most teams treat this as a vanity metric.
It’s actually a prospecting channel.
Claude Code uses a skill like github-repo-signals to turn this into a workflow. Point it at one or more repos — your competitors’, adjacent tools, or open-source projects your ICP depends on.
- It extracts every interaction type — stars, forks, issues, PRs, comments, contributions — and scores each user by engagement depth
- Users get clustered by company. When 3+ people from the same org are active on a repo, that’s not individual curiosity — that’s an account-level signal
- For each company cluster, it identifies who the actual buyer is — not just the developer who starred, but also the VP or CTO who signs purchase orders
What you get: A scored lead list organized by company, with engagement scores showing how deeply each account is already interacting with your category. A repo with 1,000 stars might surface 50–100 companies where someone on the engineering team has already evaluated what you sell.
2. Turn Hacker News and Reddit threads into qualified pipeline
Developer communities are where buying intent shows up earliest. Someone posting “what tool should I use for X” on HN is at the top of their buying journey. Someone posting “we built a hacky internal tool for Y and it’s falling apart” is describing a problem your product solves — and they’re ready to pay for a real solution.
Claude Code uses a community-signals skill to scan HN and Reddit with targeted queries:
- Generates search patterns from your product, competitors, and pain points
- Scrapes HN (free via Algolia API) and Reddit (via Apify) for matching posts and comments
- Scores each user by intent strength — “looking for alternatives” ranks higher than a passing mention
- Boosts users who appear across multiple threads or subreddits (cross-platform presence = stronger signal)
- Captures the original post or comment for each lead
What you get: A scored list sorted by intent strength, each with the exact post that triggered the signal. By the time someone shows up through inbound, they’ve been posting about the problem on HN for months. This workflow catches them earlier — when they’re evaluating, not when they’ve already decided.
3. Mine conference speaker lineups before the event happens
When a VP gets on stage and talks about “scaling our data infrastructure,” they’re publicly declaring their priorities and budget. Conference speakers are the highest-intent leads in any industry.
The best time to reach them isn’t at the event — it’s two weeks before, when they’re deep in the topic.
Claude Code uses an event-signals skill to automate this:
- Searches upcoming industry conferences and events on Luma, Meetup etc for events in your space — looking back 90 days and forward 180 days
- Scrapes speaker lineups, hosts and sponsor lists from events
- Captures each speaker’s talk title and topic — this becomes your outreach angle
- For sponsors, identifies the company and the person who likely approved the sponsorship budget
What you get: A lead list where every contact comes with a built-in conversation starter — “I saw your talk on [topic] at [event].” Specific, relevant, and proves you did your homework. One of the highest-converting cold email angles there is.
4. Find people who are actively leaving your competitors
Somewhere right now, someone is posting “looking for alternatives to [your competitor]” on Reddit.
Someone else is leaving a 2-star review on G2. A third person is commenting on your competitor’s Product Hunt launch about pricing concerns.
These are switching signals — people in active evaluation who’ve already committed budget to your category.
Claude Code uses a competitor-signals skill to monitor this across multiple channels:
- Scrapes Product Hunt commenters and upvoters on competitor launches
- Searches HN and Reddit for competitor mentions, complaints, and “alternatives to X” posts
- Pulls names from competitor case studies and testimonials — proven buyers who already said yes to the category
- Prioritizes by signal type: switching signals at the top (immediate outreach), case study companies next (account-based plays), then general engagers
What you get: A prioritized lead list segmented by urgency, with the original context attached. Your outreach references exactly what they said: “I saw your comment about [specific pain] with [competitor] — we solve exactly that.”
5. Turn incomplete inbound leads into fully enriched profiles
You get 50 demo requests this week. Half have Gmail addresses and no company name. A third have titles like “Manager” with no context. You can’t route, score, or prioritize any of them without spending an hour per lead on manual research.
The first four workflows are about finding new leads from external signals. This one is about making the leads you already have actually useful.
Claude Code uses an inbound-lead-enrichment skill to fill every gap:
- Resolves company from email domain and researches firmographics — size, industry, funding stage, tech stack, recent news
- Builds a full person profile — title, seniority, department, tenure, LinkedIn URL
- Discovers the buying committee — economic buyer, champion, evaluator, end user — so you know who else is involved in the decision
- Checks your CRM for existing relationships, active deals, and outreach history to prevent duplicate contact
What you get: Every half-empty form fill becomes a fully enriched lead record — company data, stakeholder map, seniority-based routing recommendation, and CRM context. What used to take an hour of manual research per lead now happens automatically for your entire inbound queue.
Five signals, one enrichment backbone: SixtyFour
You probably noticed something missing from each workflow above: where does the contact data actually come from?
Every workflow detects a signal and identifies the right person or company. But a signal without an email is just a name on a list. The step that turns all five workflows into actionable lead lists is the same: SixtyFour enrichment.
After each workflow runs, the AI agent calls SixtyFour’s API to resolve verified contact data for every lead:
/enrich-lead— Takes a name + company and returns email, phone, LinkedIn URL, title, seniority level, skills, and education. This is the primary endpoint every workflow calls./enrich-company— Takes a company name or domain and returns firmographic data — size, industry, tech stack, funding. Used in the GitHub and inbound workflows to understand the company before finding the right person./find-email— Targeted email discovery for specific individuals. Used when /enrich-lead returns low confidence or when the only data point is a name and website./qa-agent— Qualifies an enriched lead against your ICP criteria before it enters your outbound pipeline. Filters out bad fits so you’re only reaching out to leads worth your time.
And because SixtyFour is accessible through MCP tools, any AI agent — Claude Code, Goose, or your own — can call these endpoints directly without custom integration code.
Workflow → SixtyFour Endpoints
- GitHub → Leads: `/enrich-company`, `/enrich-lead`, `/find-email`
- Events → Leads: `/enrich-lead`, `/enrich-company`
- Competitors → Leads: `/enrich-lead`, `/enrich-company`
- Communities → Leads: `/enrich-lead`, `/find-email`
- Inbound → Enriched: `/enrich-lead`, `/enrich-company`, `/find-email`
Five different signal sources. One enrichment layer that makes all of them work.
Start with one workflow
Pick the signal you’re already seeing:
- “We have a popular GitHub repo but aren’t prospecting from it” → github-repo-signals
- “There’s a conference next month in our space” → event-signals
- “Our competitor just launched or got press” → competitor-signals
- “Our buyers hang out on HN and Reddit” → community-signals
- “Our inbound leads come in half-empty” → inbound-lead-enrichment
All five skills are free and open-source in the Goose Skills Library. Each one takes < 2 minutes to set up.
Pick one signal. Run the skill. See what comes back.