playbook-ai-specificity

Generates a personalized offer sentence tail for cold email merge fields from company descriptions.

Updated May 28, 2025
One-click install
npx skills add https://github.com/meloShaya/insteltech --skill playbook-ai-specificity-meloshaya
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: playbook-ai-specificity
Source: https://github.com/meloShaya/insteltech/tree/main/crm/library/skills/playbooks/playbook-ai-specificity
Command: npx skills add https://github.com/meloShaya/insteltech --skill playbook-ai-specificity-meloshaya

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Cold emails often read like they were sent to thousands of people because the offer is stated generically. This Skill writes the one sentence tail that restates a client's capability in the prospect's own nouns, so the email reads as written for that specific company. ## Core Features & Use Cases - Specificity Line Generation: Produces a 6-14 word merge field that slots into "I think we can help you {{specificity_line}}", starting with a plain verb and naming a real product anchor copied from the company data. - Deterministic Verification Guard: Checks that the anchor is quoted from the company evidence, the offer item is quoted from the locked client offer block, and the line passes mechanical copy rules and a Flesch-Kincaid grade 7 gate on the tail. - Abstain and Blank Handling: Returns blank for non-fit companies, treats blanks as a targeting signal, and bans spintax as the blank-handling mechanism. - Use Case: A bookkeeping firm emailing ecommerce brands gets lines like "know the landed cost of every pool float before you price it" instead of "we help with margin reporting", with Clay table and workflow builds provided. ## Quick Start Ask the AI to write a specificity line for a company by providing its domain, company name, company description, and a locked client offer block listing what the sender actually does.

Frequently Asked Questions about playbook-ai-specificity

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write a personalized cold email line for each company?

Provide the company description and a locked client offer block, then generate a 6-14 word tail that names a real product anchor from the company data and restates one offer capability in the prospect's own nouns. The line slots into "I think we can help you {{specificity_line}}".

How do I verify AI-generated personalization lines are accurate?

Use a deterministic guard that checks the anchor is copied word for word from the company evidence and the offer item is copied from the client offer list. A line built on an invented anchor or an off-list promise fails the check automatically.

What model should I use for generating merge fields in Clay?

Use gpt-4o-mini inside Clay, or a nano-class model at minimal reasoning effort in your own scripts. A reasoning model with the level unset costs roughly 19x more and returns blanks on many rows.

Why does my reading-level gate blank every generated line?

You are grading the full rendered sentence instead of the generated tail. The frame "Specifically, I think we can help you." scores 5.7 on its own, so any tail pushes the render to 7.0-9.8. Gate only the tail against grade 7.

When should a personalization line be left blank?

Blank when the company does not fit the offer, such as schools, government bodies, or charities, or when the description has no real product word to anchor on. Expect 10-20% blanks on normal lists; over 30% means the list is wrong for the campaign.

Why does fetching company homepages return empty content?

Modern ecommerce sites frequently return nothing to a plain fetch; measured results included 0 bytes, 16 bytes, 1 byte, and refused connections. Use a capped rendering proxy only when the plain fetch returns under about 400 characters.