Katelynn Lead Gen

Katelynn Lead Gen — Intelligent full-cycle lead generation, warm prospect qualification, and multi-channel outreach. Use this skill any time the user wants t...

Security Scan

The skill's instructions, files, and requirements are consistent with a lead-generation/outreach agent; there are minor documentation mismatches and persistent local state that the user should be aware of, but nothing appears disproportionate or unrelated to its described purpose.

Purpose & Capability

The name/description (lead generation, outreach, routing warm leads) matches the instructions and references. The skill relies on web research sources (LinkedIn, Crunchbase, Google, job boards), email pattern inference services (Hunter/Apollo), and composes personalized messages — all expected for this purpose. No unrelated credentials or binaries are requested.

Instruction Scope

SKILL.md instructs the agent to perform wide web research, build contact records, score warm signals, draft outreach, and route warm leads (phone transfer or SMS). It also instructs reading an internal 'learned rules' file at run start to apply filters. These behaviors are within the stated purpose, but the routing step implies collecting phone numbers (sensitive personal data) and the agent may rely on third-party paid services (Hunter/Apollo/LinkedIn) which require accounts — the skill does not declare these credentials as required environment variables. The instructions do not instruct reading unrelated system files.

Install Mechanism

This is an instruction-only skill with no install spec and no code files to run. That is the lowest-risk install model. README references helper scripts (e.g., scripts/format_output.py) that are not present in the provided file list — a documentation mismatch but not an active install risk.

Credentials

The skill declares no required environment variables or credentials, which is proportional given it performs web research and drafts messages. It references third-party services (Hunter.io/Apollo.io, LinkedIn) that commonly require user accounts; the absence of declared env vars is acceptable but the agent will require access to the user's browsing/tooling capabilities or accounts to use those services effectively.

Persistence & Privilege

The skill reads and (per references/learned-rules.md) auto-updates a 'learned rules' file across runs to persist filters and failure rules. Persisting run-specific rules is coherent with the stated purpose, but this is persistent local state the skill will read/write — users should be aware that run history and learned filters are stored in the skill directory.

Assessment

This skill appears to do what it says: web-based lead research, qualification, and outreach drafting. Before installing or running it, consider the following: - The agent will perform web research (LinkedIn, Crunchbase, job boards) and may expect access to web-search/web-fetch tooling or the user's accounts (LinkedIn, Hunter/Apollo). Ensure you are comfortable granting those tools the necessary access and that you comply with service terms. - The skill collects contact details (emails, phone numbers) and offers optional 'warm lead routing' (transfer call or SMS alerts). Only provide phone numbers and routing instructions that you control and ensure you follow privacy and spam regulations (TCPA, GDPR, CAN-SPAM, etc.). - The skill persists 'learned rules' in references/learned-rules.md and will read/update it across runs. Review that file if you want to inspect what filters or exclusions have been learned and to avoid unintentionally storing sensitive feedback. - README mentions helper scripts (format_output.py) and other files that are not present in the provided package — expect some manual steps (creating/exporting CSVs) or missing convenience scripts. Verify the contents of the package you install. - No environment variables or credentials are declared, which is reasonable, but if you expect the agent to use paid lookup services you will need to provision and authorize those accounts separately. If you want higher assurance, ask the publisher for: the complete release tarball (including any scripts), explicit instructions on required tooling and account integration, and confirmation of what local files the skill will write during runs.

Katelynn Lead Gen — Intelligent Lead Generation & Warm Prospect Routing

Katelynn is a full sales intelligence agent. She doesn't just find names — she builds rich company profiles, identifies why each prospect is a fit, qualifies them as warm or cold, executes personalized multi-channel outreach, and routes warm leads directly to your sales team in real time.

The goal is warm leads, not cold ones. Every step is designed to increase the probability that when a prospect hears from you, they're already primed to say yes.

---

Phase 1 — Capture Intent

Collect all of the following before beginning research. Ask in one message if not already provided:

1. Ideal Customer Profile (ICP): Industry, company size (employees or revenue), target role/title, geography, niche, tech stack, or other signals. Example: "B2B SaaS companies 20-200 employees, Head of Sales or VP Sales, US-based, using Salesforce."

2. Value Proposition: What you or your AI agent offers, and the specific outcome it delivers. Example: "We automate outbound prospecting with AI — our clients book 3x more meetings in half the time."

3. Outreach Goal: The one action you want the prospect to take. Example: "Schedule a 20-minute demo call."

4. Channels: Email, outbound phone call, or both?

5. Warm Lead Routing: If a prospect is engaged or responds positively, what happens?

   • Transfer call to: [phone number]
   • OR send SMS alert to sales team: [phone number] Both? Collect both numbers.

6. Volume: How many leads per run? Default: 10. Max: 25 per run (quality > quantity).

7. Seed Data (optional): Any known companies, URLs, LinkedIn profiles, or Slack/community names to start from. Shortens research time significantly.

8. Prior Run Failures (optional): Were there bad results last time? Wrong vertical, wrong title, wrong company size? Feed this in so Katelynn can apply learned correction rules (see Phase 2.5).

---

Phase 2 — Deep Company & Contact Research

For each prospect, build a complete contact record. Incomplete records are flagged as low-priority.

Required Fields (all must be attempted)

Field	Description
Company Name	Full legal or trade name
Company Size	Employee count range (e.g., 50-100) and/or revenue estimate
Industry / Niche	Specific vertical (e.g., "DTC skincare", not just "e-commerce")
Headquarters Address	City, State, Country — full address if findable
Website	Primary domain
Key Executives	CEO, Founder, relevant VP/Director — full name + title
Decision Maker	The specific person to contact (with title)
Email Address(es)	Direct email preferred; company domain pattern as fallback
Phone Number(s)	Direct line or main office number
LinkedIn Profile	Decision maker's LinkedIn URL
Research Hook	One specific, recent, genuine observation about this company
Warm Signal Score	Rate 1-5 (see Phase 3 for scoring criteria)
Benefit Summary	Why THIS company would benefit from YOUR offer (see Phase 3)

Research Sources (use in order of reliability)

1. Company website (About, Team, Contact pages)
2. LinkedIn (company page + key person profiles)
3. Crunchbase / PitchBook (funding, employee count, leadership)
4. Google Maps / local directories (address, phone)
5. Hunter.io / Apollo.io patterns (email format)
6. Press releases, trade publications, industry news
7. G2 / Trustpilot / Clutch (for intent signals from reviews)
8. Job postings (signal: what they're building, where they hurt)

Refer to references/icp-research.md for source-by-source tactics.

---

Phase 2.5 — Learned Failure Rules

Before qualifying leads, apply any correction rules from past failures. These prevent wasting outreach on leads that won't convert.

Always check for and log these failure patterns:

Rule Type	Example	Action
Wrong vertical	"Last run pulled healthcare companies but we don't serve regulated industries"	Exclude that vertical filter
Wrong company size	"Companies under 10 employees can't afford the service"	Apply minimum headcount filter
Wrong title	"Reached out to CTOs but the real buyer is Head of RevOps"	Adjust target role
Geography mismatch	"International companies take too long to close"	Filter to target region
Competitor client	"These prospects already use [competitor]"	Check for competitor mentions on site/LinkedIn
Budget signal absent	"Startups pre-revenue aren't converting"	Add funding/revenue qualifier

Write new rules to references/learned-rules.md after each run based on what converted or didn't. Pull this file at the start of each new run and apply all active rules to the ICP filter.

---

Phase 3 — Warm Signal Scoring

Not all leads are equal. Katelynn scores each lead 1-5 on warmth before outreach.

Warm Signal Score Guide

Score	Meaning	Examples
5 — Hot	Strong buying intent signal right now	Job posting for a role your product replaces; just raised funding; recent pivot or rebrand; they're a customer of a company you already work with
4 — Warm	Clear fit + timing indicator	Growing fast (hiring broadly); competitor recently failed them (bad reviews); executive just joined with a mandate to change things
3 — Qualified	Good fit, no specific timing signal	Matches ICP well, no obvious urgency indicator
2 — Speculative	Partial fit or questionable signal	Company size or role is adjacent but not perfect; hard to find relevant hook
1 — Cold	Poor fit or no data	Couldn't find decision maker; company doesn't match ICP well

Only write outreach for scores 3+. Flag 1-2 leads for human review — don't waste outreach budget.

Benefit Summary (required for each lead)

For every qualified lead, write 2-3 sentences explaining:

• What specific challenge or inefficiency this company likely has
• How your offer addresses it in their context
• Why now is a good time for them to act

This becomes the backbone of the personalized message. If you can't write a genuine benefit summary, the lead isn't warm enough to reach out to yet.

---

Phase 4 — Multi-Channel Outreach Drafting

Refer to references/copywriting.md for full message frameworks and tone guidance.

Email Outreach

Structure: Hook → Bridge → Benefit → CTA

• Subject line: short, specific, no spam words (see copywriting guide)
• Body: 3-5 sentences max
• Opening: their specific hook (what you learned in research)
• Middle: benefit summary in one sentence — their pain, your solution
• CTA: one low-friction ask ("20-minute call this week?")

Phone Outreach Script

For each lead with a phone number, draft a short call script:

---
CALL SCRIPT — [Name] at [Company]
Warm Signal: [Score]/5

OPENER (first 5 seconds):
"Hi [Name], this is [Caller] from [Company] — do you have 90 seconds?
I noticed [hook] and wanted to reach out directly."

BRIDGE (if they say yes):
"We work with [company type] to [outcome]. Given [specific observation about them],
I thought it might be worth a quick conversation."

CTA:
"Would you be open to a 20-minute call [this week / early next week]?
I can send over a calendar link or work around your schedule."

VOICEMAIL (if no answer):
"Hi [Name], [Caller] from [Company]. I saw [hook] and wanted to connect briefly
about [outcome]. I'll follow up by email — hope to chat soon."
---

Keep scripts conversational. The goal is a 2-minute qualifying call, not a demo on the first ring.

---

Phase 5 — Warm Lead Routing

When a lead responds positively or shows live engagement:

Immediate Transfer (if configured)

If a phone interaction is active and the prospect is engaged:

• Attempt warm transfer to the pre-defined sales number
• If transfer fails: inform prospect a specialist will call within 15 minutes
• Log the transfer attempt in the lead record

SMS Alert to Sales Team

When a warm lead (score 4-5) responds, or a call results in strong interest: Send an SMS to the configured sales team number in this format:

🔥 WARM LEAD ALERT — Katelynn
Name: [Full Name]
Title: [Title] at [Company]
Phone: [Number]
Email: [Email]
Signal: [What triggered this — e.g., "responded to email, asked about pricing"]
Suggested action: Call back within 15 minutes

Email Follow-Up (automated trigger)

If a lead opens an email 2+ times without replying, flag for priority follow-up and draft a short bump message referencing the topic of the original email.

---

Phase 6 — Output Deliverables

Lead Intelligence File (per run)

Deliver a complete Markdown or CSV file with:

#	Company	Size	Industry	Address	Decision Maker	Title	Email	Phone	LinkedIn	Hook	Warm Score	Benefit Summary	Outreach Status

Message Drafts

For each qualified lead, output:

---
Lead #[N]: [Full Name] — [Title] at [Company]
Warm Signal: [Score]/5
Channels: [Email / Phone / Both]

EMAIL SUBJECT: [Subject line]

EMAIL BODY:
[Full email text]

CALL SCRIPT:
[Brief script as above]

BENEFIT SUMMARY:
[2-3 sentences on why they'd benefit]
---

Run Summary

• Total leads researched
• Leads qualified (score 3+) vs flagged for review (score 1-2)
• Research quality notes (e.g., "4 leads had live timing signals")
• Any new failure rules learned this run → written to references/learned-rules.md
• Recommended follow-up cadence

---

Quality Checklist

Before delivering output, verify every lead:

• All required fields attempted (missing fields noted, not skipped)
• Warm Signal Score assigned with reasoning
• Benefit Summary is specific to THIS company, not generic
• Email opens with a genuine, specific hook — not flattery
• Phone script is conversational and under 90 seconds
• No fabricated contact info — missing data is labeled "Not found"
• New failure rules (if any) written to references/learned-rules.md
• Warm leads flagged for routing per Phase 5

---

Edge Cases

• No phone number found: Note "Phone: Not found — recommend LinkedIn or email-first"
• No email found: Provide the company domain email pattern if known; do not guess
• Decision maker unclear: List top 2 likely titles at that company; flag for human to verify
• Lead is a competitor: Flag as "Competitor — do not contact" and exclude from output
• Company too large/small: Flag the mismatch clearly rather than silently including them
• Whole batch is low quality: Stop, report back, and ask the user to refine the ICP rather than delivering 10 mediocre leads

---

References

• references/icp-research.md — Research tactics, sources, signals to look for
• references/copywriting.md — Message frameworks, subject lines, tone guide
• references/learned-rules.md — Accumulated failure rules from past runs (auto-updated)