Social lead gen

Social Lead Gen

Find people who need your product — from what they're actually saying on social media.

Unlike traditional lead gen tools that search company databases, this skill finds high-intent leads from live conversations. It discovers people actively expressing the problems your product solves across Twitter, Instagram, and Reddit — powered by 1.5B+ indexed posts via Xpoz MCP.

---

⚡ Prerequisites

1. Xpoz MCP must be configured and authenticated. Follow the xpoz-setup skill.
2. Web search and web fetch tools must be available (included with OpenClaw).

Verify Xpoz is ready:

mcporter call xpoz.checkAccessKeyStatus

If not hasAccessKey: true, follow xpoz-setup first, then return here.

---

How It Works

┌──────────────────┐    ┌──────────────────┐    ┌──────────────────┐
│  PHASE 1: LEARN  │ →  │ PHASE 2: SEARCH  │ →  │ PHASE 3: SCORE   │
│                  │    │                  │    │                  │
│ Research product │    │ Generate queries  │    │ Score by intent  │
│ Analyze website  │    │ Search Twitter    │    │ Rank leads       │
│ Find competitors │    │ Search Instagram  │    │ Write outreach   │
│ Map pain points  │    │ Search Reddit     │    │ Export results   │
│ Validate w/ user │    │ Poll for results  │    │ Track sent leads │
└──────────────────┘    └──────────────────┘    └──────────────────┘
     (one-time)             (repeatable)            (repeatable)

---

Phase 1: Product Research (One-Time Setup)

This phase builds deep context about the user's product. Run it once; the results are stored and reused.

Step 1: Ask the user for a reference

Ask the user:

"What product or service do you want to find leads for? Give me a reference — a website URL, GitHub repo, product description, or anything that describes what you offer."

The user may provide:

• A website URL (e.g., https://example.com)
• A GitHub repo (e.g., https://github.com/org/repo)
• A product name + description
• Multiple references

Step 2: Deep research

For each reference provided, gather as much context as possible:

For websites:

• Use web_fetch to read the homepage, pricing page, about page, docs
• Use web_search to find reviews, mentions, comparisons, press coverage

For GitHub repos:

• Use web_fetch to read the README
• Check stars, description, topics
• Use web_search for mentions, discussions, blog posts about it

For product names:

• Use web_search to find the product website, reviews, alternatives
• Then fetch and analyze the results

Extract and organize:

{
  "product": {
    "name": "",
    "website": "",
    "tagline": "",
    "description": "",
    "category": "",
    "pricing": ""
  },
  "value_proposition": {
    "what_it_does": "",
    "key_features": [],
    "differentiators": []
  },
  "target_audience": {
    "primary_icp": "",
    "segments": [
      {
        "name": "",
        "description": "",
        "pain_points": [],
        "where_they_hang_out": {
          "subreddits": [],
          "hashtags": [],
          "communities": []
        }
      }
    ]
  },
  "pain_points_solved": [],
  "competitors": [
    {
      "name": "",
      "website": "",
      "how_different": ""
    }
  ],
  "social_proof": {
    "testimonials": [],
    "case_studies": [],
    "notable_customers": []
  },
  "keywords": {
    "product_terms": [],
    "pain_point_terms": [],
    "competitor_terms": [],
    "industry_terms": []
  }
}

Step 3: Validate with the user

Present your findings in a clear summary:

"Here's what I learned about your product:

[Product Name] — [tagline]

What it does: [description]

Target audience: [segments]

Pain points you solve:

• [pain point 1]
• [pain point 2]

Competitors: [list]

Key differentiators: [list]

Does this look right? Anything I should add or correct?"

WAIT for the user to confirm or correct. Do not proceed until they approve.

If the user corrects something, update the profile and re-validate.

Step 4: Generate search queries

Based on the validated profile, generate targeted search queries for each platform:

Query categories:

1. Pain point queries — People expressing the problems the product solves
2. Competitor frustration queries — People complaining about competitors
3. Tool/solution seeking queries — People actively looking for what the product offers
4. Industry discussion queries — People in the target audience discussing relevant topics

For each query, specify:

• Platform (Twitter, Instagram, Reddit)
• Query string (using boolean operators)
• Target subreddits (for Reddit)
• Minimum engagement thresholds
• Lookback period

Generate 4-6 queries per platform (12-18 total).

Step 5: Store the profile

Save the validated profile and generated queries:

mkdir -p data/social-lead-gen
# Save product-profile.json and search-queries.json

Present the generated queries to the user:

"I've generated [N] search queries across Twitter, Instagram, and Reddit. Here are a few examples:

Twitter (pain points): "[query example]" Reddit (tool seeking): "[query example]" in r/[subreddit]

Ready to search for leads?"

---

Phase 2: Lead Discovery (Repeatable)

Run this phase whenever you want fresh leads. Uses the stored profile and queries.

Step 1: Load profile

cat data/social-lead-gen/product-profile.json
cat data/social-lead-gen/search-queries.json

If these files don't exist, run Phase 1 first.

Step 2: Execute searches

For each generated query, call the appropriate Xpoz MCP tool:

Twitter:

mcporter call xpoz.getTwitterPostsByKeywords \
  query="GENERATED_QUERY" \
  startDate="LOOKBACK_DATE" \
  limit=50 \
  fields='["id","text","authorUsername","likeCount","retweetCount","replyCount","impressionCount","createdAtDate"]'

Instagram:

mcporter call xpoz.getInstagramPostsByKeywords \
  query="GENERATED_QUERY" \
  startDate="LOOKBACK_DATE" \
  limit=50

Reddit:

mcporter call xpoz.getRedditPostsByKeywords \
  query="GENERATED_QUERY" \
  startDate="LOOKBACK_DATE" \
  limit=50

Always poll for results:

mcporter call xpoz.checkOperationStatus operationId="OPERATION_ID"

Poll every 5 seconds until status is completed.

Step 3: Find people (not just posts)

For high-engagement posts, also search for the people behind them:

mcporter call xpoz.getTwitterUsersByKeywords \
  query="GENERATED_QUERY" \
  limit=50

This finds users who frequently post about the topic — potential repeat customers or influencers.

---

Phase 3: Lead Scoring & Output

Scoring Framework (1-10)

Score each lead based on signals from the product profile:

Signal	Points	Example
Explicitly asking for a solution	+3	"Can anyone recommend a [product category]?"
Complaining about a competitor	+2	"[Competitor] is too expensive / broken / limited"
Has a project blocked by the pain point	+2	"I need [capability] but can't find a good tool"
Active in target community	+1	Posts in relevant subreddits / uses relevant hashtags
High engagement on the post	+1	>10 likes or >5 comments
Recent post (< 48 hours)	+1	Time-sensitive opportunity
Profile matches ICP	+1	Developer, marketer, researcher — matches target segment
Selling a competing solution	-3	They're a competitor, not a lead
Irrelevant context	-2	Mentioned keyword but in unrelated context

Tiers:

• Tier 1 (Score 8-10): Hot leads — high intent, act fast
• Tier 2 (Score 6-7): Warm leads — worth engaging
• Tier 3 (Score 5): Watchlist — monitor for future intent
• Below 5: Skip

Deduplication

Before reporting any lead, check against previously sent leads:

cat data/social-lead-gen/sent-leads.json

Key format: {platform}:{author}:{post_id}

After reporting, add to the tracking file.

Output Format

For each lead, provide:

1. Who: Username, platform, profile summary
2. What they said: Direct quote from their post
3. Post URL: Clickable link
   • Twitter: https://twitter.com/{username}/status/{id}
   • Reddit: https://www.reddit.com/r/{subreddit}/comments/{id}/
   • Instagram: https://www.instagram.com/p/{shortcode}/
4. Score: X/10 with reasoning
5. Why they're a fit: Connection between their pain and your product
6. Suggested outreach: Pre-written reply customized to their situation
7. Engagement: Likes, comments, shares
8. Posted: Date and relative time

Outreach Guidelines

When writing suggested replies:

• Reference their exact situation from the post
• Mention specific features of the product that solve their problem
• Keep it conversational, not salesy
• Be honest — if the product doesn't perfectly fit, say so
• Always include a disclosure if the user is affiliated with the product

Example:

"I had the same problem! Ended up using [Product] — it does [specific capability they need]. [Brief differentiator]. Worth checking out: [URL]

(Disclosure: I work with [Product])"

---

Updating the Profile

If the user's product evolves, they can re-run Phase 1:

"My product has changed — we now also support [X]. Can you update the profile?"

Re-run the research, validate, and regenerate queries. The stored profile is overwritten.

---

Example Session

User: "Find leads for my product. Here's our website: https://example.com"

Agent:

1. Fetches and analyzes example.com (homepage, pricing, docs)
2. Searches for reviews, competitors, mentions
3. Presents findings: "Here's what I learned..."
4. User confirms: "Yes, but we also target enterprise customers"
5. Agent updates profile, generates 15 search queries
6. User says "Go find leads"
7. Agent runs queries across Twitter/Instagram/Reddit
8. Returns: "Found 12 leads — 3 Tier 1, 5 Tier 2, 4 Tier 3"
9. Each lead has URL, score, quote, outreach copy

Next day: "Find more leads" → Agent loads saved profile, runs fresh queries, deduplicates against yesterday's results.

---

Responsible Use

• Respect platform terms of service for Twitter, Instagram, and Reddit
• Don't spam — outreach should be genuine and helpful
• Disclose affiliations in any outreach messaging
• Respect privacy — only use publicly available information
• Quality over quantity — 5 great leads beat 50 mediocre ones

---

Resources

• Xpoz: xpoz.ai — social intelligence MCP powering the searches
• Setup: xpoz-setup on ClawHub — one-time auth
• Search reference: xpoz-social-search on ClawHub — full search patterns

---

Built for ClawHub • Powered by Xpoz