A complete example showing how to use RUNE specifications with agents.md architecture for a multi-agent code quality system.
agents.md| File | Description |
|---|---|
agents.md |
Agent system architecture with 2 agents |
tools/doc_generator.rune |
RUNE spec for docstring generation tool |
tools/test_validator.rune |
RUNE spec for test coverage validation tool |
Generates Python docstrings in Google, NumPy, or Sphinx styles. Analyzes function signatures, parameters, and body logic to produce accurate documentation.
Tool: generate_docstring (spec)
Validates test coverage by comparing source code against test files. Reports coverage percentage, identifies missing tests, and provides actionable recommendations.
Tool: validate_test_coverage (spec)
agents.md defines the agent architecture – who does what.rune file that defines the contract – how it works# Generate a docstring
claude "Read tools/doc_generator.rune and generate a Google-style docstring for this function: def add(a: int, b: int) -> int: return a + b"
# Validate test coverage
claude "Read tools/test_validator.rune and validate the test coverage of my_module.py"
agents.md and both .rune files to project knowledge# Generate the doc_generator tool implementation
claude "Generate a Python implementation from tools/doc_generator.rune"
# Generate the test_validator tool implementation
claude "Generate a Python implementation from tools/test_validator.rune"
To add a new agent or tool:
agents.md with its role, capabilities, and tools.rune file for each tool in tools/tool_name --> [spec](tools/tool.rune)