The prospecting skill runs one five-phase workflow for building prospect lists: define the ICP, discover candidates, qualify each with sources, score and tier them, and output a lead sheet. It forks into four branches: SaaS, general B2B, local small business and demand-signal discovery for early-stage teams.
Each lead carries evidence, a confidence level and a verified date. Emails are checked for deliverability before they reach the list, and a set of compliance guardrails rules out bulk scraping, bypassing logins and unprovenanced data.
What You Get
- A lead sheet as a chat table or CSV with score, signal, contact, email status, sources and confidence
- An ICP statement with pass and fail criteria before any discovery starts
- Hot, Warm, Cold and Skip scoring plus account tiers that set research depth
- A top outreach targets list with one sentence of rationale for each lead
- A list of open questions for anything the agent could not verify
Common Use Cases
- Building a first outbound list for a SaaS product from funding, hiring and tech stack signals
- Finding local businesses by category and area with website status noted
- Locating early design partners from public pain signals in forums and reviews
- Cleaning and verifying an existing list before a cold email sequence
What Makes It Unique
It splits prospecting into four branches, each with its own signals, and adds a demand-signal branch that scores evidence of pain from cited public sources. The compliance rules come first in the file: no bulk scraping of LinkedIn or Google Maps, define audiences in Sales Navigator and pull contacts from licensed data, and keep source URL and date for every contact.
Under The Hood
The skill is an instruction file with ten reference documents on branches, data sources, enrichment, signal plays, account research and compliance. It has no scripts of its own. The agent works with whatever tools the user has: browser-assisted research on public pages, enrichment and verification services such as Apollo, Clay, Hunter or Firecrawl through their APIs or MCP servers, and it may write a CSV file when the list is long. It is told to space requests and never scrape platforms in bulk.