How does it optimize tool usage?

It provides templates to calibrate tool triggering, making agents more proactive with 'default-to-action' or more conservative with 'default-to-research' instructions, while maximizing parallel tool calls.

Can I control the agent's verbosity?

Yes, the skill includes verbosity control patterns that allow you to toggle between concise progress updates and detailed summaries.

Does this help with long-running or autonomous tasks?

Absolutely. It includes structured strategies for context management, state tracking via JSON/text files, and prescriptive methods for resuming work across fresh context windows.

What is the benefit for frontend developers?

It helps avoid generic 'AI slop' aesthetics by enforcing unique typography, cohesive themes, and sophisticated motion rather than predictable, cookie-cutter layouts.

How does this skill improve Claude's coding accuracy?

It provides specific instructions to investigate files before answering, avoid over-engineering, and focus on general-purpose solutions rather than just passing test cases.

Expert Prompt Engineering

Name: Expert Prompt Engineering
Author: lbussell

bylbussell

0•

Developer Tools

Optimizes LLM agent behavior and output quality through advanced instruction techniques, tool calibration, and state management patterns.

This skill provides a comprehensive framework for writing effective LLM prompts and optimizing agentic workflows within Claude Code. It establishes core principles like explicitness and context-setting while offering specialized instructions for formatting control, tool usage calibration, and state management for long-running tasks. By implementing these patterns, users can ensure their agents are proactive yet minimal, avoid common hallucinations, and produce high-quality frontend designs that avoid generic 'AI slop' aesthetics. It is particularly useful for developers looking to refine how Claude interacts with complex codebases and external tools.

Key Features

01State management patterns for long-running complex tasks

02Explicit instruction and context-setting frameworks

03Minimalist agentic coding and anti-hallucination guardrails

04Creative frontend design and aesthetic guidelines

05Advanced tool usage and parallel execution logic

060 GitHub stars

Use Cases

01Optimizing agent instructions for multi-step autonomous coding tasks

02Refining LLM response formatting for specific text-to-speech or UI requirements

03Controlling agent proactivity and tool-triggering sensitivity in sensitive codebases

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lbussell/agent-skills prompt-engineering

For use in Claude.ai and ChatGPT

Download Skill