Add new command skill.pike-rag for Knowledge Graph Augmented Generation

[giuseppe-trisciuoglio]

Add new command skill.pike-rag for Knowledge Graph Augmented Generation
Background
Standard Retrieval-Augmented Generation (RAG) often struggles with complex, multi-hop queries that require global knowledge across multiple documents. The PIKE-RAG (Photorealistic and Interactive Knowledge Engine for RAG) framework addresses this by utilizing Hyper-Relational Knowledge Graphs (HKGs) to capture intricate relationships between entities, significantly improving reasoning and information retrieval for specialized domains.
Feature Description
Create a new skill command skill.pike-rag.md in the developer-kit-skills plugin. This skill will enable the AI to process documents into structured hyper-relational formats, allowing for more precise information extraction and complex query answering than traditional vector-search methods.
Workflow Position
This skill is utilized during the Context Gathering and Implementation phases of a task to ensure the AI has a deep, graph-based understanding of the technical documentation.
Key Responsibilities of the New Command

• Document Parsing & HKG Construction:
  • Extract entities and attributes from provided documentation.
  • Establish "hyper-relational" links (links between links or multi-entity relationships) rather than simple triplets.
• Schema Alignment:
  • Ensure the generated knowledge graph adheres to the domain-specific schema required for the project.
• Multi-Hop Retrieval:
  • Enable the model to traverse the graph to answer questions that require connecting information from disparate parts of the documentation.
• Evidence-Based Generation:
  • Generate code or documentation where every claim is backed by a specific node or path within the constructed HKG.
• Conflict Resolution:
  • Identify and flag contradictory information found across different source documents during the graph synthesis phase.
    Acceptance Criteria
• New skill definition file plugins/developer-kit-skills/skills/skill.pike-rag.md is created.
• The skill includes specific prompts for HKG construction from raw text/markdown.
• The skill provides logic for Multi-Hop Reasoning to resolve complex technical queries.
• Implementation follows the core PIKE-RAG principles: capturing interactive relationships and structured knowledge engine capabilities.
• Documentation is provided on how to trigger this skill during a specs.task-implementation flow.
• A test suite is provided to verify that the skill can correctly identify a 3-hop relationship from a sample dataset.