These guidelines enable the AugmentCode Local Agent to effectively utilize the AGOR multi-agent development coordination platform. The agent should read AGOR documentation from workspace sources and follow structured development protocols.
When working on development tasks, you are operating within the AGOR (AgentOrchestrator) framework - a sophisticated multi-agent development coordination platform. Your primary responsibilities:
- Read AGOR Documentation: Always start by reading the AGOR protocol files from workspace sources
- Use AGOR Dev Tools: Always use the programmatic dev tools for all AGOR operations
- Follow AGOR Protocols: Use structured workflows, development tools, and coordination methods
- Create Snapshots: Always create snapshots before ending sessions using AGOR dev tools
- Maintain Context: Use AGOR's memory and coordination systems for session continuity
- Commit Frequently: Use quick_commit_and_push() often to keep work synchronized
MANDATORY: Execute these functions to initialize the AGOR system:
python3 -c "
import sys
sys.path.insert(0, 'src')
from agor.tools.dev_tools import get_agor_initialization_guide, get_available_functions_reference, test_all_tools
print('🚀 AGOR INITIALIZATION')
print('=' * 50)
print(get_agor_initialization_guide())
print('\n📋 AVAILABLE FUNCTIONS REFERENCE')
print('=' * 50)
print(get_available_functions_reference())
print('\n🧪 TESTING TOOLS')
print('=' * 50)
test_all_tools()
print('✅ AGOR initialization complete!')
"
If the above functions don't work, read these files as fallback:
- src/agor/tools/README_ai.md (role selection and initialization)
- src/agor/tools/AGOR_INSTRUCTIONS.md (comprehensive operational guide)
- src/agor/tools/index.md (documentation index for efficient lookup)
🔧 Dev Tools (Primary Interface):
- Always use AGOR dev tools for all operations
- Never manually create or manage .agor files
- All memory operations handled automatically by dev tools
📸 Snapshot System:
- Snapshots saved to
.agor/agents/{agent_id}/snapshots/on memory branches - Use
create_development_snapshot()function only - Never manually create snapshot files
🧠 Memory Branch Architecture:
- Memory branches (e.g.,
agor/mem/main) store all .agor files - Working branches never contain .agor directories
- Cross-branch commits handled automatically by dev tools
You will typically operate as a 🔍 Worker Agent for:
- Codebase analysis and exploration
- Feature implementation and debugging
- Technical documentation and code explanation
- Direct development work and task execution
Note: Multi-agent coordination strategies are experimental and not recommended for production use. Focus on the mature snapshot/memory system and dev tools.
If working on external projects (where AGOR is installed separately from the project):
# ALWAYS use external integration system instead of direct imports
from agor.tools.external_integration import get_agor_tools
# Initialize with automatic AGOR detection
tools = get_agor_tools()
tools.print_status() # Check integration status
# Use AGOR functions through the tools object
tools.generate_pr_description_output("content")
tools.create_development_snapshot("title", "context")Why this is critical: Direct imports fail when AGOR is installed separately. The external integration system provides automatic detection and fallback mechanisms.
See: src/agor/tools/EXTERNAL_INTEGRATION_GUIDE.md for complete setup instructions.
📸 Snapshot & Memory Management:
create_development_snapshot(title, context, next_steps)- Create comprehensive work snapshotsgenerate_session_end_prompt(task_description, brief_context)- Generate handoff promptsgenerate_pr_description_output(content)- Create PR descriptions (brief content only)generate_release_notes_output(content)- Create release notes (brief content only)
⚡ Quick Development Workflow:
quick_commit_and_push(message, emoji)- USE FREQUENTLY - Efficient commit and pushget_workspace_status()- Check project and git statustest_all_tools()- Verify all dev tools work correctly
🧠 Memory Operations (Automatic):
- All memory operations happen automatically on separate memory branches
- Never manually create or edit .agor files
- Snapshots automatically saved to
.agor/agents/{agent_id}/snapshots/on memory branches
For External Projects (Recommended):
from agor.tools.external_integration import get_agor_tools
tools = get_agor_tools()
# Use functions through tools object:
tools.create_development_snapshot("title", "context", ["next", "steps"])
tools.quick_commit_and_push("message", "🔧")
tools.generate_pr_description_output("content")For AGOR Development Only:
# Only use direct imports when working ON AGOR itself
from agor.tools.dev_tools import (
create_development_snapshot,
quick_commit_and_push,
generate_pr_description_output,
)🚨 MANDATORY Session End Process:
- Use quick_commit_and_push() frequently during development to keep work synchronized
- Create development snapshot using
create_development_snapshot()function - Generate handoff prompt using
generate_session_end_prompt()function - Never manually create .agor files - all memory management is automatic
📍 Snapshot Storage (Automatic):
- Snapshots automatically saved to
.agor/agents/{agent_id}/snapshots/on memory branches - Never manually interact with .agor directories - use dev tools only
- Memory branches (e.g.,
agor/mem/main) handle all coordination files automatically
End each response with suggestions like:
- "Would you like me to create a development snapshot and handoff prompt for this work?"
- "I can generate PR notes using the dev tools if you're ready to create a pull request."
- "Let me know if you'd like me to commit our progress using quick_commit_and_push()."
Remember: Use quick_commit_and_push() often - it's efficient, saves actions, and keeps your branch synchronized when multiple users/agents are working on the project.
CRITICAL: All snapshots must capture comprehensive user context like a transcriptionist:
What to Capture:
- Every detailed explanation the user provides about their reasoning and decision-making
- Strategic thinking and priorities - their long-term vision and how current work fits
- Technical preferences - their architectural philosophy and implementation choices
- User voice and reasoning patterns - maintain their thought process and decision rationale
- Complete context behind decisions - why certain choices were made, not just what was done
Example of GOOD context capture: "User emphasized that multi-agent coordination strategies are largely untested and need more work, preferring to focus on the snapshot/memory branch system and dev tools as the more developed features. They want comprehensive context capture like a transcriptionist, including all reasoning and decision-making process. The user specifically noted that agents should never manually interact with snapshots - all memory management must be done using dev tools. They also emphasized the importance of quick_commit_and_push() functionality for efficiency and branch synchronization."
Example of POOR context capture: "Updated user guidelines and fixed some issues."
- Use real git commands (not simulated)
- Use quick_commit_and_push() frequently for efficient workflow
- Push changes regularly for backup and collaboration
- Pattern:
quick_commit_and_push("Descriptive message", "🔧")instead of manual git commands
- Keep files under 500 lines when creating new projects
- Use modular, testable code structure
- No hard-coded environment variables
- Maintain clean separation of concerns
- Write comprehensive tests using TDD approach
- Document code with clear comments and docstrings
- Follow security best practices
- Optimize for maintainability and extensibility
- Role Selection: src/agor/tools/README_ai.md
- Complete Guide: src/agor/tools/AGOR_INSTRUCTIONS.md
- Documentation Index: src/agor/tools/index.md
- External Integration: src/agor/tools/EXTERNAL_INTEGRATION_GUIDE.md (CRITICAL for external projects)
- Snapshot Guide: src/agor/tools/SNAPSHOT_SYSTEM_GUIDE.md
- Strategy Guide: docs/strategies.md
- Development Guide: docs/agor-development-guide.md (when working on AGOR itself)
- Bundle Mode: docs/bundle-mode.md
- Standalone Mode: docs/standalone-mode.md
- Usage Guide: docs/usage-guide.md
- Quick Start: docs/quick-start.md
- NEVER end a session without creating a snapshot - Use
create_development_snapshot()function - NEVER manually create .agor files - All memory management is automatic via dev tools
- Use quick_commit_and_push() frequently - Keep work synchronized and prevent loss
- Snapshots are automatically saved - To
.agor/agents/{agent_id}/snapshots/on memory branches - Always use dev tools functions - Never manually interact with AGOR memory system
- Test your work - Verify functionality before completion
- Follow AGOR protocols precisely - Read documentation and use programmatic functions
You are successfully using AGOR when you:
- ✅ Initialize AGOR using programmatic dev tools functions
- ✅ Use quick_commit_and_push() frequently during development
- ✅ Create snapshots using create_development_snapshot() function only
- ✅ Never manually interact with .agor directories or files
- ✅ Use dev tools for all AGOR operations (snapshots, handoffs, PR descriptions)
- ✅ Follow structured development practices with frequent commits
- ✅ Provide comprehensive context through proper dev tools usage
- Use AGOR's feedback tools to provide feedback on AGOR itself
- Suggest improvements to workflows and documentation
- Report issues or exceptional scenarios
- Help evolve AGOR protocols based on real-world usage
Remember: AGOR transforms AI assistants into sophisticated development coordinators. Your adherence to these protocols ensures effective coordination, context preservation, and successful project outcomes.