ICP Radar

Lead scoring and morning brief agent. Pulls recent contacts from your CRM, scores each one against your ICP definition using Claude, and delivers a prioritized list with recommended actions — plus a spoken brief to start your day.

What this skill does

1. Fetches the 20 most recent contacts from HubSpot CRM
2. Scores each contact 0–100 against your ICP criteria from CONFIG.md
3. Returns top 3 priority leads with fit level and recommended next action
4. Generates a spoken morning brief summarizing the pipeline

Configuration

Before running, set in CONFIG.md:

HUBSPOT_API_KEY=
YOUR_ICP_INDUSTRY=          # e.g. B2B SaaS, fintech, professional services
YOUR_ICP_COMPANY_SIZE=      # e.g. 10-200 employees
YOUR_ICP_TITLES=            # e.g. CMO, VP Marketing, Head of Growth, Founder
YOUR_ICP_GEOGRAPHY=         # e.g. Canada, North America
YOUR_ICP_SIGNALS=           # e.g. AI tool adoption, recent funding, hiring for marketing
YOUR_NAME=                  # Used in spoken brief greeting
ANTHROPIC_API_KEY=

How to invoke

# CLI
agent run icp-radar

# Voice (after ACTIVATION.md setup)
"[your wake word] leads"

# Scheduled — set your preferred time in crontab or Task Scheduler
# See ACTIVATION.md for scheduling instructions

ICP scoring model

Each contact is scored 0–100 across five dimensions:

Score interpretation:

• 75–100 = HIGH fit — prioritize immediately
• 50–74 = MEDIUM fit — nurture and qualify
• Below 50 = LOW fit — deprioritize or disqualify

Claude system prompt (what runs under the hood)

You are a lead scoring analyst. Score each of the following CRM contacts 
against this ICP definition:

Industry: [YOUR_ICP_INDUSTRY]
Company size: [YOUR_ICP_COMPANY_SIZE]
Decision maker titles: [YOUR_ICP_TITLES]
Geography: [YOUR_ICP_GEOGRAPHY]
Buying signals to look for: [YOUR_ICP_SIGNALS]

For each contact, return:
- name, company, title
- icp_score (0-100)
- fit_level (HIGH / MEDIUM / LOW)
- signal_alignment: what signals match your ICP
- next_best_action: one specific recommended action
- reasoning: 1 sentence explaining the score

Return top 3 contacts only, sorted by icp_score descending.
Also return a spoken_brief: a 60-second spoken morning brief starting with 
"Good morning. Here is your lead intelligence for today."
Return valid JSON only.

Output format

{
  "top_leads": [
    {
      "name": "Contact name",
      "company": "Company name",
      "title": "Job title",
      "icp_score": 87,
      "fit_level": "HIGH",
      "signal_alignment": "B2B SaaS, CMO title, 50-employee company",
      "next_best_action": "Send personalized outreach referencing their recent product launch",
      "reasoning": "Strong ICP match on all dimensions — decision maker at target company size."
    }
  ],
  "spoken_brief": "Good morning. Here is your lead intelligence for today..."
}

Chaining with other skills

# Turn top lead insight into a LinkedIn post
agent run icp-radar | agent run linkedin-post-generator

# Natural language
agent ask "Score my leads then write a LinkedIn post about what I found"

Extending to other CRMs

This skill is written for HubSpot. To adapt for another CRM:

• Replace the HubSpot API endpoint with your CRM's contacts endpoint
• Update the API key configuration accordingly
• The Claude scoring step works the same regardless of CRM source

Related skills

• linkedin-post-generator — turn lead insights into content
• ai-news-digest — combine with news for a full morning briefing
• See ACTIVATION.md for voice and scheduled setup