Skip to content

Amplihack Agents

This directory contains specialized agents that extend Claude Code's capabilities for the amplihack framework.

Recently Added Agents (2025-10-27)

Knowledge Work Agents

These agents were ported from amplifier to enhance amplihack's knowledge processing and complexity management capabilities:

1. knowledge-archaeologist

Value: HIGH - Traces evolution of knowledge and identifies valuable abandoned approaches

  • Use when: Understanding concept evolution, paradigm shifts, discovering old solutions for new problems
  • Key capability: Temporal analysis of knowledge - how ideas evolve, decay, and resurrect
  • Output: Temporal layers, lineage trees, paradigm shifts, decay patterns, revival candidates

2. concept-extractor

Value: MEDIUM - Extracts structured knowledge from documents

  • Use when: Processing articles, papers, or documents for knowledge synthesis
  • Key capability: Identifies atomic concepts, relationships, tensions, and uncertainties
  • Output: Structured JSON with concepts, relationships, tensions, uncertainties

3. insight-synthesizer

Value: MEDIUM - Discovers revolutionary connections and breakthrough insights

  • Use when: Stuck on complex problems, seeking innovative solutions, need unexpected connections
  • Key capability: Collision-zone thinking, pattern-pattern recognition, simplification cascades
  • Output: Collision experiments, cross-domain patterns, simplifications, revolutionary insights

Integration with Amplihack Philosophy

These agents complement amplihack's existing capabilities:

  • knowledge-archaeologist + analyzer: Understand code evolution and why patterns were chosen
  • concept-extractor + knowledge-builder: Enhance documentation and knowledge base creation
  • insight-synthesizer + optimizer: Find breakthrough simplifications and novel solutions

For ambiguity handling and post-task cleanup, use the production agents in amplihack/specialized/:

  • amplihack/specialized/ambiguity - Handles complex requirements with multiple valid interpretations
  • amplihack/specialized/cleanup - Ensures code quality and philosophy compliance post-implementation

Usage Examples

Using knowledge-archaeologist for code understanding:

User: "Why did we choose this specific architecture pattern?"
Claude: "Let me invoke the knowledge-archaeologist agent to trace the evolution of this pattern and document the reasoning"

Using concept-extractor for documentation:

User: "Process these design documents and extract the key concepts"
Claude: "I'll use the concept-extractor agent to build a structured knowledge base from these documents"

Using insight-synthesizer for innovation:

User: "We need a breakthrough approach to this performance bottleneck"
Claude: "Let me deploy the insight-synthesizer agent to explore revolutionary connections and find simplification cascades"

Agent Interaction Patterns

Sequential Pattern (Common)

  1. concept-extractor → Extract knowledge from documents
  2. insight-synthesizer → Find revolutionary connections
  3. architect → Design implementation
  4. builder → Implement solution
  5. cleanup (amplihack/specialized) → Ensure quality

Parallel Pattern (When Appropriate)

  • ambiguity (amplihack/specialized) + architect → Handle complex requirements simultaneously
  • knowledge-archaeologist + analyzer → Understand code and history in parallel

Iterative Pattern (For Exploration)

  1. insight-synthesizer → Generate multiple approaches
  2. ambiguity (amplihack/specialized) → Preserve tensions between approaches
  3. architect → Design solution that accommodates multiple paths
  4. builder → Implement flexible architecture

Philosophy Alignment

All agents strictly follow amplihack's core principles:

  • Simplicity: Start simple, add only justified complexity
  • Modular: Self-contained modules with clear interfaces
  • Working code: No stubs or dead code
  • Test-driven: Tests before implementation

The new agents enhance these principles by:

  • Preserving valuable complexity (use amplihack/specialized/ambiguity)
  • Learning from past simplifications (knowledge-archaeologist)
  • Enforcing simplicity post-implementation (use amplihack/specialized/cleanup)
  • Structuring knowledge clearly (concept-extractor)
  • Finding breakthrough simplifications (insight-synthesizer)

Contributing New Agents

When adding new agents:

  1. Follow the agent template format (see existing agents)
  2. Define clear use cases with examples
  3. Specify tool requirements
  4. Document output format
  5. Align with amplihack philosophy
  6. Add to this README with value assessment
  7. Update main README agent count

Testing Agents

Test new agents with:

# Test agent loads correctly
claude --agent knowledge-archaeologist

# Test agent with sample task
claude --agent concept-extractor "Extract concepts from this architecture document"

See Also