This SOP guides business users through creating reusable "skills" for Claude without writing code. A skill is defined as a three‑part construct—name, description/trigger, and step‑by‑step instructions—that captures a repetitive job in plain English. The guide walks users through a seven‑step workflow: (1) pick a suitable repetitive task (with examples for coaches, agencies, and general businesses), (2) write down the job using a five‑question template (outcome, trigger, input, output format, rules/voice), (3) feed this information to Claude via a copy‑paste prompt that asks Claude to ask clarifying questions and then generate a skill file, (4) test the skill on real, messy business inputs, (5) iteratively refine the skill by giving plain‑English corrections, (6) save the finalized skill in the Claude account for repeated use, and (7) repeat the process weekly to build a library of personalized automation tools. The SOP emphasizes that skills are built like SOPs, that early feedback dramatically improves quality, and includes a homework assignment to build the first skill within the week.
Topics
Skill Definition: Three‑Part Structure (Name, Trigger, Instructions)
Explains that a skill consists of a name, a description that serves as the trigger, and detailed step‑by‑step instructions written in plain English.
Choosing the Right Repetitive Task
Guidelines for selecting tasks that are frequent, have consistent execution, and clear rules, with concrete examples for coaches, agencies, and any business.
Job Specification Template: Five Key Questions
Provides a structured set of five questions (what should it do, when should it kick in, what goes in, what should come out, what are the rules) to capture the job in plain English before prompting Claude.
Prompting Claude to Build the Skill
A copy‑paste prompt formula that tells Claude the job details, asks for clarifying questions, and returns a ready‑to‑save skill file.
Testing with Real Business Inputs
Emphasizes using authentic, messy inputs to evaluate the skill’s output, highlighting that realistic testing is the only true validation.
Iterative Refinement: Early Feedback Loop
Describes how to give plain‑English corrections (e.g., shorten, adjust voice) over a few cycles to lock in quality, noting that early corrections have outsized impact.
Saving, Deploying, and Scaling Skills
Instructions for saving the finalized skill in Claude, using it repeatedly, and building a habit of creating one skill per week to compound automation benefits.