Epic: Subagent Foundation #2
Description
Description
This epic focuses on extending the core functionality with a subagent system that enables specialized tasks within the dev-agent ecosystem. The Subagent Foundation provides a framework for coordinating multiple AI-driven agents that assist with code analysis, planning, and context provision.
Architecture: Context provision and read-only analysis agents that work together to reduce AI hallucinations by providing rich, accurate context.
Goals
- ✅ Create a robust subagent coordinator for managing multiple specialized agents
- ✅ Implement a communication protocol between subagents
- ✅ Develop the planner subagent for task breakdown and organization
- ✅ Build the explorer subagent for pattern discovery and code exploration
- ✅ Create the GitHub context subagent for indexing and searching issues/PRs
Acceptance Criteria
- ✅ Subagent coordinator successfully manages the lifecycle of multiple agents
- ✅ Communication protocol allows efficient message passing between agents
- ✅ Planner subagent can analyze GitHub issues and break them down into actionable tasks
- ✅ Explorer subagent identifies patterns, similar code, and relationships across repositories
- ✅ GitHub context subagent indexes and searches GitHub issues/PRs, linking them to code
- ✅ All components are thoroughly tested with >80% test coverage (achieved 100% for utilities!)
- ✅ Documentation is comprehensive and clear
Definition of Done
- ✅ All code passes linting, formatting, and type checking
- ✅ All tests are passing with >= 80% coverage (557 tests passing, 100% utils coverage)
- ✅ Documentation is complete, including API reference and user guides
- ✅ Real-world dogfooding: Used Planner to break down Issue Implement Planner Subagent #8, Explorer for code discovery
- ✅ Agents successfully coordinate through message passing
Completed Work
Coordinator (Issue #7)
- Message routing and agent lifecycle management
- Context sharing between agents
- Health checks and graceful shutdown
- 17 integration tests
Planner Subagent (Issue #8)
- Analyzes GitHub issues via
ghCLI - Breaks issues into actionable tasks with effort estimates
- CLI command:
dev plan <issue-number> - 65 tests (50 utils + 15 integration)
Explorer Subagent (Issue #9)
- Pattern discovery across codebase
- Similar code identification
- Relationship mapping
- CLI commands:
dev explore pattern/similar/relationships - 123 tests (90 utils + 33 integration)
GitHub Context Subagent (Issue #10)
- Indexes GitHub issues, PRs, discussions
- Semantic search with relevance scoring
- Links GitHub items to code files
- CLI commands:
dev gh index/search/context - 61 tests (47 parser + 14 integration)
Future Work
Automation Layer (Future Epic #3):
The PR subagent originally planned for this epic has been deferred to a future automation-focused epic. Our current focus is on context provision (read-only analysis) rather than automation (write operations).
Future automation capabilities could include:
- PR creation and management
- Automated code fixes
- Issue triage and labeling
- See Issue Implement Context API and GitHub Integration #5 for GitHub automation planning
Impact
- Total Tests: 557 passing (30 test files)
- Utility Coverage: 100% for all subagents
- Documentation: 4 comprehensive READMEs (2,000+ lines)
- Lines Added: ~8,000 (including tests and docs)
- Real-world validation: Successfully dogfooded during development
Priority: High (Core foundation for agent system)