Ruthless concision for AI prompts and context - "Not One Word Wasted"
Ernest Hemingway wrote with brutal efficiency. Every word earned its place. No word wasted. This skill applies that ruthlessness to AI prompts, context documents, and instructions.
Why: AI context is expensive. Every wasted word costs tokens, slows comprehension, dilutes signal. Concision is not just style—it's engineering efficiency.
Anti-inspiration: Charles Dickens. Verbose exposition. Excessive adjectives. Meandering sentences. Everything we avoid.
Wrong: "The system will be using a database that has been configured to store user data." Right: "System stores user data in database."
Cut: actually, basically, essentially, generally, literally, really, very, quite, just, simply, clearly, obviously.
Wrong: "We basically just need to simply verify that the user is actually authenticated." Right: "Verify user authentication."
Wrong: "Facilitate the implementation of a solution for managing authentication state." Right: "Manage authentication state."
Wrong: "Each individual user has their own personal preferences." Right: "Users have preferences."
Wrong: "The authentication system validates user credentials and manages session state while also handling token refresh and expiration, plus it logs all authentication attempts for security auditing purposes."
Right: "Authentication system validates credentials, manages sessions, refreshes tokens, logs attempts."
Skip pleasantries, apologies, hedging:
Wrong: "The system needs to handle user registration, and it also needs to manage user authentication, as well as dealing with password resets, not to mention email verification."
Right: System handles:
I would like you to please help me by analyzing this codebase in order to
identify any potential areas where we might be able to improve the overall
performance characteristics of the system. Specifically, I'm interested in
understanding whether there are any obvious bottlenecks that could be causing
slowdowns, or if there are any particular functions that seem to be taking
longer than they should to execute.
Analyze codebase for performance bottlenecks. Identify slow functions.
This skill is designed to provide comprehensive assistance with the task of
implementing test-driven development practices in your codebase. It will help
guide you through the process of writing tests before you write your actual
implementation code.
Write tests first, then code. Red-green-refactor cycle.
Commands should tell Claude exactly what to do. No preamble. No explanation of why commands exist.
Wrong:
This command helps you create a new feature by first analyzing the requirements
and then helping you design the architecture before you start implementation.
Right:
Analyze requirements. Design architecture. Create implementation plan.
CLAUDE.md documents provide context. Context should be dense, scannable, structured.
Wrong:
This module is responsible for handling all of the various aspects related to
user authentication within the system, including things like login, logout,
session management, and token handling.
Right:
Handles authentication: login, logout, sessions, tokens.
Workflow phases should state purpose and output. No marketing copy.
Wrong:
This is an interactive and collaborative design session where we'll work
together to explore the architectural possibilities and come up with the best
possible approach for your feature.
Right:
Interactive design session. Output: architecture decisions, component breakdown.
But these are rare. Default to ruthless.
When writing new content:
When editing existing content:
When other skills reference this: They should apply these principles automatically when generating prompts, instructions, or documentation.
Good hemingwayesque writing:
Bad writing (what we eliminate):
Not one word wasted. Every word earns its place. Cut ruthlessly. Write tight.
Hemingway would approve.