Prompt Builder

Overview

prompt-builder provides a systematic workflow for creating effective prompts in Claude Code skills. It applies prompt engineering principles, uses proven templates, and validates quality to ensure prompts are clear, specific, actionable, and produce reliable results.

Purpose: Create high-quality prompts for skills, workflows, tasks, and operations

Pattern: Workflow-based (5-step process)

Key Benefit: Transforms vague instructions into precise, effective prompts that Claude can execute reliably

When to Use

Use prompt-builder when:

• Creating new skill instructions
• Writing workflow steps that Claude will execute
• Defining task operations
• Building automation prompts
• Improving existing prompts that produce inconsistent results
• Ensuring prompts follow best practices

Prerequisites

Before building prompts:

• Clear objective: Know what the prompt should accomplish
• Context understanding: Understand the situation where the prompt will be used
• Success criteria: Define what "good output" looks like

Prompt Building Workflow

Step 1: Understand Context

Before writing any prompt, understand the full context.

What to Identify:

1. Goal: What should this prompt accomplish?

   • Create something new?
   • Transform existing content?
   • Analyze and provide insights?
   • Make a decision?

2. Audience: Who will use this prompt?

   • Claude directly (in skill)?
   • Developer building skills?
   • End user interacting with system?

3. Situation: When/where will this prompt be used?

   • Part of sequential workflow?
   • Independent task operation?
   • Conditional branch?
   • Error recovery?

4. Constraints: What limitations exist?

   • Time limits?
   • Output format requirements?
   • Tools available/unavailable?
   • Dependencies on other steps?

5. Success criteria: How do you know it worked?

   • Measurable outcome?
   • Specific format?
   • Validation criteria?

Questions to Ask:

• What problem does this prompt solve?
• What does "success" look like?
• What could go wrong?
• What information does Claude need?
• What should Claude NOT do?

Example Context Analysis:

Goal: Have Claude write a comprehensive guide for a reference file
Audience: Claude (executing skill)
Situation: Step 3 of 6 in skill-building workflow
Constraints: Must be <5,000 words, specific format
Success: Complete guide covering all topics, properly formatted

Common Context Mistakes:

• ❌ Skip context analysis, jump straight to writing
• ❌ Assume context is obvious
• ❌ Don't define success criteria
• ❌ Ignore constraints

Best Practices:

• ✅ Write down context explicitly
• ✅ Identify all constraints upfront
• ✅ Define measurable success
• ✅ Consider edge cases

→ Output: Context analysis document

→ Next: With context clear, define the specific task

Step 2: Define Task Clearly

Transform the goal into a clear, specific, actionable task definition.

Task Definition Elements:

1. Action Verb: What specific action?

   • Good: "Write", "Analyze", "Create", "Transform", "Extract"
   • Avoid: "Handle", "Deal with", "Work on", "Process"

2. Object: What is being acted upon?

   • Be specific: "the authentication module" not "the code"
   • Include type: "markdown file", "Python function", "data structure"

3. Constraints: What limits or requirements?

   • Format: "as a bulleted list", "in JSON format"
   • Length: "under 500 words", "10-15 items"
   • Style: "imperative voice", "technical language"

4. Quality Criteria: What makes output good?

   • Completeness: "covering all edge cases"
   • Accuracy: "matching the specification exactly"
   • Usability: "beginner-friendly explanations"

5. Context References: What information to use?

   • "based on the examples in references/"
   • "following the pattern from Step 2"
   • "using the template below"

Task Clarity Checklist:

• Action verb is specific and unambiguous
• Object is clearly identified
• Constraints are explicitly stated
• Success criteria are measurable
• Context references are provided
• Can be completed without additional questions

Good vs Bad Task Definitions:

❌ Bad: "Update the documentation"

• Vague action, unclear object, no criteria

✅ Good: "Write a Getting Started section for README.md covering installation, basic usage, and first example. Use imperative voice, keep under 300 words, include 2-3 code examples."

• Clear action, specific object, defined constraints and criteria

❌ Bad: "Make the code better"

• Subjective, unmeasurable, no direction

✅ Good: "Refactor the authentication function to use async/await pattern, add error handling for network failures, and include JSDoc comments for each parameter."

• Specific improvements, clear criteria, measurable outcome

Task Definition Template:

[ACTION VERB] [SPECIFIC OBJECT] that [QUALITY CRITE

...