* feat(agents): implement category-based agent organization system - Organize agents into domain categories (core, development, content, data, product, learning) - Move 10 agents to category subdirectories with proper git rename tracking - Update 13 subagents with category and type metadata in frontmatter - Add category metadata files (0-category.json) documenting common patterns - Implement local registry fallback in install script for offline development - Add comprehensive validation suite with 15 automated tests (100% pass rate) - Enhance registry validation with duplicate ID and consistency checks - Update eval framework with intelligent path resolution (backward compatible) - Archive legacy eval structure to _archive/ for reference - Update all documentation to reflect category-based structure - Bump version to 0.5.0 with accurate CHANGELOG Technical Details: - 23 agents organized (10 category agents, 13 subagents) - 6 category directories created - Path resolution supports both agent IDs and category paths - Registry schema updated to v2.0.0 - 159 files changed, 872 insertions(+), 10738 deletions(-) BREAKING CHANGE: Agent file paths now use category structure. Update references from .opencode/agent/openagent.md to .opencode/agent/core/openagent.md. Eval framework maintains backward compatibility via path resolution. * refactor(evals): consolidate documentation and enhance test infrastructure - Remove temporary project tracking files (PHASE_5_COMPLETE, PROJECT_COMPLETE, etc.) - Consolidate evaluation framework docs into main README - Enhance test execution with improved logging and multi-prompt support - Move system-builder from core to meta category - Add comprehensive test suites for openagent with organized structure - Create evaluation test structure for all subagents - Clean up archived workflows and redundant documentation - Update registry to reflect new agent organization - Add shared test templates and golden test patterns * feat(ci): add manual trigger support for bot-created PRs in validate-registry workflow - Add pr_number input to workflow_dispatch for manually triggering validation - Fetch PR details dynamically when triggered manually - Support both automatic (PR event) and manual (workflow_dispatch) triggers - Enable validation of bot-created PRs like automated version bumps - Update branch detection and push logic to handle both trigger types - Add documentation explaining how to manually trigger for bot PRs * feat(evals): add explicit context file validation to test framework - Add expectedContextFiles field to test YAML schema for explicit context file specification - Enhance context-loading-evaluator to support both auto-detect and explicit validation modes - Update documentation with comprehensive guide and examples - Clean up archived legacy test structure (90+ old test files) - Add new example tests demonstrating explicit context validation - Backward compatible with existing tests * feat(evals): add multi-agent logging system and performance optimizations Implement comprehensive multi-agent logging and performance improvements for the eval framework. Task 01: Multi-Agent Logging System - Add complete hierarchical logging module (evals/framework/src/logging/) - SessionTracker: tracks parent-child delegation hierarchies - MultiAgentLogger: pretty-prints logs with visual indentation - Formatters: box characters and emoji formatting - 37 passing unit tests (session-tracker, logger, integration) - Integrate with SDK event stream handler - Hook into session.created, message.updated, message.part.updated events - Real-time child session detection via timestamp heuristics - Message deduplication for cleaner output - Enable in debug mode only (<1% performance overhead) - Add demo script and comprehensive documentation Task 02: Performance Optimizations - Reduce grace period from 5s to 2s (67% reduction, 10-20% faster tests) - Add PerformanceMetricsEvaluator for bottleneck identification - Collects tool latencies, inference time, idle time - Provides full performance visibility - Update smoke test to validate delegation (core feature) - Replace simple read test with multi-agent delegation test - Tests both parent and child agent functionality - Validates multi-agent logging system Results: - 37 unit tests passing - Smoke test passing (95/100 score) - Clean hierarchical logging output - 10-20% faster test execution - Full multi-agent visibility Files changed: 16 files, +2678 lines Time saved: ~6-8 days (completed in 6 hours vs 3-5 day estimate) * feat(evals): show child agent execution in non-debug mode - Enable MultiAgentLogger in both debug and non-debug modes - Add verbose parameter to control output level - Non-verbose mode shows concise child session lifecycle: - Child agent started message - Child agent completed with duration - Verbose mode (--debug) shows full delegation hierarchy - Update documentation to reflect new behavior This provides visibility into delegation without overwhelming output, giving confidence that child agents are actually running. * feat(evals): improve delegation testing and behavior validation Major improvements to eval framework: Test Suite: - Add 00-smoke-test.yaml (basic read operation) - Rename 01-smoke-test.yaml → 02-delegation-test.yaml (delegation test) - Add simple-responder test agent for delegation testing - Add debug test: simple-subagent-call.yaml Evaluators: - Add agent-model-evaluator.ts for agent/model validation - Enhance behavior-evaluator.ts with detailed task tool output - Improve delegation-evaluator.ts with better evidence tracking Schema & Execution: - Add expectedAgent and expectedModel to test schema - Improve test executor with better logging and error handling Documentation: - Update evals/README.md with new features and performance improvements - Remove outdated MULTI_AGENT_LOGGING_COMPLETE.md Registry: - Add simple-responder test agent to registry Fixes: - Fix dashboard.sh path resolution These changes provide better visibility into delegation, improved test coverage, and clearer validation of agent behavior. * chore: add GitHub Actions workflow and plugin documentation - Add .github/workflows/evals/run-evaluations.yml for automated eval testing - Add dev/ai-tools/opencode/plugins/Plugin-inspiration.md for plugin development reference
11 KiB
OpenCoder - Specialized Development Agent
Your expert development partner for complex coding tasks
Table of Contents
- What is OpenCoder?
- When to Use OpenCoder
- When to Use OpenAgent Instead
- Core Capabilities
- Workflow
- Multi-Language Support
- Code Standards
- Subagent Delegation
- Examples
- Tips for Best Results
What is OpenCoder?
OpenCoder is a specialized development agent focused on complex coding tasks, architecture analysis, and multi-file refactoring. It follows strict plan-and-approve workflows with modular and functional programming principles.
Key Characteristics:
- 🎯 Specialized - Deep focus on code quality and architecture
- 🔧 Multi-language - Adapts to TypeScript, Python, Go, Rust, and more
- 📐 Plan-first - Always proposes plans before implementation
- 🏗️ Modular - Emphasizes clean architecture and separation of concerns
- ✅ Quality-focused - Includes testing, type checking, and validation
When to Use OpenCoder
✅ Multi-file refactoring (4+ files)
- Refactoring an entire module
- Implementing patterns across multiple components
- Restructuring architecture
✅ Architecture analysis and improvements
- Analyzing codebase structure
- Identifying architectural issues
- Proposing design improvements
✅ Complex code implementations
- Features spanning multiple modules
- Implementations requiring > 60 minutes
- Features with complex dependencies
✅ Pattern discovery and application
- Finding existing patterns in codebase
- Implementing consistent patterns
- Refactoring to match established patterns
✅ Deep codebase analysis
- Understanding complex code flows
- Documenting architecture
- Identifying technical debt
When to Use OpenAgent Instead
Use openagent for:
- ❓ Questions about code or concepts
- 📝 Documentation tasks
- 🔄 Simple 1-3 file changes
- 🎯 General workflow coordination
- 💬 Exploratory conversations
Rule of thumb:
- OpenAgent = General coordinator (questions, docs, coordination, simple tasks)
- OpenCoder = Development specialist (complex coding, architecture, refactoring)
Core Capabilities
Code Implementation
- Modular architecture design
- Functional programming patterns
- Type-safe implementations
- SOLID principles adherence
- Clean code practices
- Proper separation of concerns
Analysis & Review
- Architecture analysis
- Pattern discovery
- Code quality assessment
- Technical debt identification
- Performance analysis
Refactoring
- Multi-file refactoring
- Pattern implementation
- Architecture improvements
- Code modernization
- Technical debt reduction
Quality Assurance
- Type checking (TypeScript, Python, Go, Rust)
- Linting (ESLint, Pylint, etc.)
- Build validation
- Test execution
- Incremental validation
Workflow
Phase 1: Planning (Required)
OpenCoder always proposes a plan first:
1. Analyzes the request
2. Creates step-by-step implementation plan
3. Presents plan to user
4. Waits for approval
No implementation happens without approval.
For features spanning multiple modules or estimated > 60 minutes, OpenCoder delegates to @task-manager to create atomic subtasks.
Phase 2: Implementation (After Approval)
Implements incrementally - one step at a time:
For each step:
1. Implement the code
2. Run type checks (if applicable)
3. Run linting (if configured)
4. Run build checks
5. Execute relevant tests
6. Validate results
7. Move to next step
Validation happens continuously, not just at the end.
For simple subtasks, delegates to @subagents/coder-agent to save time.
Phase 3: Completion
When implementation is complete:
1. Final validation
2. User approval
3. Handoff recommendations for:
- @tester (if tests needed)
- @documentation (if docs needed)
- @reviewer (if security review needed)
Multi-Language Support
OpenCoder adapts to the project's language automatically:
TypeScript/JavaScript
- Runtime:
node,bun, ordeno - Type checking:
tsc - Linting:
eslint - Testing:
jest,vitest,mocha
Python
- Runtime:
python - Type checking:
mypy - Linting:
pylint,flake8 - Testing:
pytest,unittest
Go
- Build:
go build - Linting:
golangci-lint - Testing:
go test
Rust
- Build:
cargo check,cargo build - Linting:
clippy - Testing:
cargo test
Code Standards
OpenCoder follows these principles:
Modular Architecture
- Clear module boundaries
- Single responsibility principle
- Loose coupling, high cohesion
- Dependency injection where appropriate
Functional Patterns
- Pure functions where possible
- Immutable data structures
- Declarative over imperative
- Function composition
Type Safety
- Strong typing (when language supports)
- Explicit types over inference (when clearer)
- Type guards and validation
- Null safety
Clean Code
- Meaningful names
- Small, focused functions
- Minimal, high-signal comments
- Avoid over-complication
- Follow language conventions
Subagent Delegation
OpenCoder coordinates with specialized subagents:
@task-manager
When: Features spanning 4+ files or > 60 minutes
Purpose: Break down into atomic subtasks
Output: Task files under tasks/subtasks/{feature}/
@coder-agent
When: Simple, focused implementation tasks Purpose: Quick code implementation Output: Implemented code following specifications
@tester
When: Tests needed for implementation Purpose: Write comprehensive test suites Output: Unit, integration, and e2e tests
@reviewer
When: Security or quality review needed Purpose: Code review, security analysis Output: Review report with recommendations
@build-agent
When: Build validation needed Purpose: Type checking, build verification Output: Build status, error reports
@documentation
When: Comprehensive documentation needed Purpose: Generate API docs, guides Output: Structured documentation
Examples
Example 1: Multi-File Refactoring
opencode --agent opencoder
> "Refactor the authentication module to use dependency injection across all 8 files"
# OpenCoder will:
# 1. Analyze current structure (8 files)
# 2. Propose refactoring plan
# 3. Wait for approval
# 4. Delegate to @task-manager (8 files > 4 file threshold)
# 5. Implement subtasks one at a time
# 6. Validate incrementally
# 7. Complete when all subtasks done
Example 2: Architecture Analysis
opencode --agent opencoder
> "Analyze the architecture of this codebase and suggest improvements"
# OpenCoder will:
# 1. Scan codebase structure
# 2. Identify patterns and anti-patterns
# 3. Propose architectural improvements
# 4. Wait for approval
# 5. Implement approved changes
# 6. Validate with build and tests
Example 3: Pattern Implementation
opencode --agent opencoder
> "Implement the repository pattern for all database access across the data layer"
# OpenCoder will:
# 1. Identify all database access points
# 2. Design repository interface
# 3. Propose implementation plan
# 4. Wait for approval
# 5. Delegate to @task-manager (multi-file)
# 6. Implement repositories incrementally
# 7. Update all consumers
# 8. Add tests via @tester
# 9. Validate complete implementation
Example 4: Complex Feature
opencode --agent opencoder
> "Implement user authentication with JWT, refresh tokens, and role-based access control"
# OpenCoder will:
# 1. Analyze requirements (complex, multi-file)
# 2. Design authentication architecture
# 3. Propose implementation plan (multiple phases)
# 4. Wait for approval
# 5. Delegate to @task-manager (create subtasks)
# 6. Implement Phase 1: JWT infrastructure
# 7. Implement Phase 2: Refresh token mechanism
# 8. Implement Phase 3: RBAC system
# 9. Coordinate with @tester for test coverage
# 10. Coordinate with @reviewer for security review
# 11. Validate end-to-end
Tips for Best Results
1. Be Specific About Scope
Good: "Refactor the API layer to use dependency injection in controllers and services" Bad: "Make the code better"
2. Provide Context
If refactoring existing code, mention:
- Number of files involved
- Current patterns being used
- Desired end state
- Any constraints (can't change X, must maintain Y)
3. Review Plans Carefully
OpenCoder will show you the plan before implementation. Take time to:
- Verify the approach makes sense
- Check that all files are included
- Ensure edge cases are considered
- Request changes if needed
4. Let OpenCoder Delegate
Don't manually delegate to subagents. OpenCoder knows when to:
- Use @task-manager for breakdown
- Call @tester for tests
- Use @reviewer for security
- Leverage @coder-agent for simple tasks
5. Use Test-Driven Development
If a tests/ directory exists, OpenCoder will:
- Write tests first (when appropriate)
- Validate against tests continuously
- Ensure comprehensive coverage
6. Trust Incremental Implementation
OpenCoder implements one step at a time, not all at once. This:
- Catches errors early
- Allows for course correction
- Maintains working code between steps
- Makes debugging easier
7. Language-Specific Conventions
OpenCoder adapts to your language:
- For TypeScript: Functional, type-first approach
- For Python: Pythonic patterns, type hints
- For Go: Idiomatic Go, interfaces
- For Rust: Ownership, traits, Result types
Configuration
OpenCoder is configured in .opencode/agent/core/opencoder.md. Default settings:
temperature: 0.1 # Deterministic, precise
tools: read, edit, write, grep, glob, bash, patch
permissions:
bash: Limited (ask for risky commands)
edit: Deny secrets, node_modules, .git
Comparison: OpenAgent vs OpenCoder
| Aspect | OpenAgent | OpenCoder |
|---|---|---|
| Primary Use | General coordinator | Development specialist |
| Best For | Questions, docs, coordination | Complex coding, architecture |
| Coding Tasks | Simple (1-3 files) | Complex (4+ files) |
| Delegation | Delegates coding to opencoder | Delegates testing, review |
| Expertise | Broad, adaptive | Deep, technical |
| User Profile | Everyone (default) | Developers |
| Plan Detail | High-level | Implementation-level |
| Validation | Basic | Comprehensive (type, lint, build, test) |
Summary
OpenCoder is your specialized development partner for:
- ✅ Complex multi-file coding tasks
- ✅ Architecture analysis and improvements
- ✅ Pattern implementation and refactoring
- ✅ Deep technical implementations
Use OpenAgent for general tasks and coordination. Use OpenCoder when you need deep development expertise.
Start here:
opencode --agent opencoder
> "Your complex coding task..."
Learn more:
- OpenAgent Guide - General tasks and coordination
- Agent System Blueprint - Architecture patterns
- Research-Backed Design - Why it works