Skip to content

SKILL.md

Latest commit

 

History

History
107 lines (83 loc) · 4.04 KB

SKILL.md

File metadata and controls

107 lines (83 loc) · 4.04 KB
name rag-query-optimizer
description Automatically analyzes RAG pipeline retrieval results and generates optimized query refinement suggestions to improve answer relevance.

rag-query-optimizer

1. Name & Description

rag-query-optimizer — Automatically analyzes RAG pipeline retrieval results and generates optimized query refinement suggestions to improve answer relevance.

2. Core Capabilities

  • Analyze RAG pipeline retrieval results for relevance patterns and gaps
  • Detect low-relevance retrievals through statistical score analysis
  • Identify query-document semantic mismatches
  • Generate intelligent query refinement suggestions based on retrieval analysis
  • Suggest synonym expansions and contextual rewrites
  • Recommend query decomposition strategies for complex queries
  • Support multiple RAG pipeline configurations and retrieval strategies
  • Provide CLI interface for direct query optimization tasks
  • Integrate seamlessly with OpenClaw agent workflows
  • Handle errors gracefully with non-zero exit codes on failure

3. Triggers

  • "I need to optimize my RAG queries"
  • "Analyze my retrieval results and suggest query improvements"
  • "Help me refine my RAG pipeline queries for better relevance"
  • "Generate query optimization suggestions for my RAG system"
  • "My RAG retrieval needs improvement, what query refinements do you suggest?"
  • "Review my RAG pipeline and provide query refinement recommendations"
  • "Automate query refinement analysis for our RAG system"
  • "How can I improve my retrieval relevance scores?"
  • "Suggest better queries for my RAG pipeline"
  • "Optimize search queries for my document retrieval system"

4. Out of Scope

  • Does not execute actual RAG pipelines or perform retrievals
  • Does not implement query refinement (only generates suggestions)
  • Does not provide vector database management or indexing
  • Does not handle document preprocessing or chunking
  • Does not offer real-time query monitoring or dashboarding
  • Does not support proprietary RAG framework integrations
  • Does not provide performance benchmarking tools
  • Does not manage embedding model selection or training

5. References

  • references/rag-patterns.md — Common RAG optimization patterns for analysis matching
  • references/query-techniques.md — Query refinement methodologies used by the suggestion engine
  • references/config-examples/ — Sample configurations for different RAG systems

6. Key Files

  • SKILL.md — Skill documentation and metadata
  • README.md — User guide and usage documentation
  • scripts/main.js — CLI entry point and orchestration logic
  • scripts/analyzer.js — RAG retrieval analysis engine
  • scripts/suggester.js — Query refinement suggestion generator
  • scripts/config.js — Configuration management and validation
  • scripts/error-handler.js — Error handling and exit management
  • scripts/utils.js — Shared utility functions
  • references/rag-patterns.md — Optimization patterns reference
  • references/query-techniques.md — Refinement methodologies reference
  • assets/examples/ — Sample input/output datasets
  • assets/templates/ — Configuration templates

7. API Usage

# Analyze a retrieval results file
node scripts/main.js analyze <results-file>

# Analyze with custom configuration
node scripts/main.js analyze <results-file> --config <config-file>

# Output as JSON
node scripts/main.js analyze <results-file> --format json

# Show help
node scripts/main.js --help

8. Integration

Agent workflow integration:

  1. Agent sends retrieval results via sessions_send
  2. Skill deployed via sessions_spawn with task "optimize-query"
  3. Results returned through message tool to user's session
# Example agent invocation
node scripts/main.js analyze results.json --format json

9. Error Handling

Exit Code Meaning
0 Success — suggestions generated
1 General error — unexpected failure
2 Invalid input — malformed results file
3 Configuration error — invalid config
4 No results — empty retrieval data