$TLOGZ Prompt Flow
~/claude/claude-knowledge-base-builder.md
$catclaude-knowledge-base-builder.md

Claude Knowledge Base Builder

Create structured knowledge bases with Claude Projects, organizing documents, concepts, and relationships for consistent AI-assisted work.

Best
claude-sonnet-4
Good
gpt-4o, claude-haiku
Limited
gemini-2.5-pro, gpt-4o-mini
Updated
2026-05-22
workflow

You are a knowledge management architect. Design a structured knowledge base optimized for Claude Projects.

Domain: {{domain}} Content Sources: {{contentSources || "Existing documentation and expert input"}} Knowledge Structure: {{structure || "Hierarchical with cross-references"}} Intended Users: {{users || "Internal team"}}

Knowledge Base Architecture

1. Taxonomy Design

Create a category hierarchy:

Domain: {{domain}} ├── Category 1: [name] │ ├── Subcategory 1.1 │ ├── Subcategory 1.2 │ └── Subcategory 1.3 ├── Category 2: [name] │ ├── Subcategory 2.1 │ └── Subcategory 2.2 └── Category 3: [name] ├── Subcategory 3.1 └── Subcategory 3.2

2. Document Templates

For each content type, define a template:

Concept Document

# Concept: [Name] ## Definition ## Key Attributes ## Relationships ## Examples ## Common Misconceptions

Process Document

# Process: [Name] ## Trigger ## Steps (numbered) ## Decision Points ## Expected Outcome ## Failure Modes

Reference Document

# Reference: [Name] ## Specifications ## API / Interface ## Configuration ## Changelog

3. Cross-Reference System

Link related knowledge with typed relationships:

  • is-a: Hierarchical parent
  • part-of: Component relationship
  • depends-on: Dependency mapping
  • related-to: Soft association
  • contradicts: Conflict to resolve

4. Claude Optimization

Structure files for Claude's 200K context:

  • Chunking: Split large documents into 10-20KB files
  • Naming: category_subcategory_topic.md convention
  • Index: A root _index.md mapping all documents
  • Priority: Put most-frequently-used content first

5. Update Workflow

Maintain knowledge base quality:

  • Review Cadence: Monthly accuracy checks
  • Deprecation: Archive outdated documents with redirects
  • Versioning: Keep major revision history in each doc
  • Feedback Loop: Track what users search for most

6. Example Entries

Provide 5 example knowledge base entries following the templates above, relevant to {{domain}}.

Output with bold headers, code for file naming, --- for document separators, and clear | table | formatting for taxonomy mappings.

variables
^Enter
guide
how to use
  • Open the Claude Knowledge Base Builder workflow in your AI chat interface.
  • Replace the variables in [brackets] with your specific inputs.
  • For best results, use claude-sonnet-4 as the target model.
  • Review the generated output and iterate by refining your inputs.
  • Save your final result and share it with your team.
best use cases
  • Quickly generate claude-specific content with structured prompts.
  • Standardize claude workflows across your team using a shared template.
  • Onboard new team members with a repeatable claude process.
  • Automate claude tasks with AI-powered claude workflows.
  • Automate knowledge-base tasks with AI-powered claude workflows.
  • Automate documentation tasks with AI-powered claude workflows.
examples
  • Use Claude Knowledge Base Builder to create a claude project from scratch.
  • Adapt Claude Knowledge Base Builder for a different claude domain with custom variables.
  • Combine Claude Knowledge Base Builder with other workflows in the claude category for a complete pipeline.
  • Run Claude Knowledge Base Builder with multiple AI models to compare output quality.
  • Schedule Claude Knowledge Base Builder as a recurring claude task.
variations
  • Simplified version: remove optional variables for faster results.
  • Advanced version: add custom validation steps after generation.
  • Batch version: run Claude Knowledge Base Builder on multiple inputs sequentially.
  • claude-focused variant: emphasize claude best practices in the prompt.
  • knowledge-base-focused variant: emphasize knowledge-base best practices in the prompt.
common mistakes
  • Skipping variable customization — always replace [bracketed] placeholders.
  • Using the wrong AI model tier for complex outputs.
  • Not iterating on the first result — refinement improves quality significantly.
  • Ignoring claude best practices when customizing the prompt.
  • Using claude-sonnet-4 outside its optimal use case for this workflow.
Remix