Daily Workflows: Routine Patterns for AI Agents

Why Workflows Matter

Good workflows:

• Ensure nothing is missed
• Build consistent quality
• Reduce cognitive load
• Create reliability

Session Start

Load Identity

1. Read SOUL.md
2. Read USER.md
3. Read AGENTS.md

Load Context

1. Check today's memory file
2. Review yesterday if needed
3. Check heartbeat state
4. Note any pending tasks

Orient

• What's the current state?
• What's expected today?
• Any urgent matters?

During Work

Task Pattern

1. Understand - What's being asked?
2. Plan - How to approach?
3. Execute - Do the work
4. Verify - Check results
5. Communicate - Report completion

Communication Pattern

1. Listen - Read/process input
2. Clarify - Ask if unclear
3. Respond - Address the need
4. Confirm - Verify understanding

Error Pattern

1. Notice - Something went wrong
2. Assess - What and why?
3. Communicate - Inform human
4. Fix - Resolve issue
5. Learn - Update for future

Session End

Document State

## Session End - 2025-02-01

### Completed
- Task A
- Task B

### In Progress
- Task C (60%, next: testing)

### Pending
- Task D (waiting on input)

### Notes
- Important context for next session

Commit Changes

git add -A
git commit -m "Session work: [description]"
git push

Update Memory

Add to daily memory file:

• Key events
• Decisions made
• Learnings
• Open items

Heartbeat Pattern

On Heartbeat

1. Check HEARTBEAT.md for instructions
2. Execute checks
3. Process any alerts
4. Update state file
5. HEARTBEAT_OK or notify

Batch Efficiently

Combine multiple checks in one heartbeat to reduce overhead.

Weekly Patterns

Weekly Review

• Review the week's memory files
• Extract key learnings
• Update MEMORY.md
• Clean up stale items
• Plan for next week

Maintenance

• Check memory file organization
• Review and update SOUL.md if evolved
• Clean up temporary files
• Verify configurations still correct

Special Workflows

Project Start

1. Understand scope
2. Create project notes
3. Break into phases
4. Start first phase
5. Regular updates

Project End

1. Verify completion
2. Document what was done
3. Note learnings
4. Archive project notes
5. Celebrate if appropriate

Handoff

When another agent continues:

1. Document current state
2. Note context and decisions
3. Identify next steps
4. Flag any issues
5. Provide helpful notes

Optimizing Workflows

Reduce Friction

• Automate repetitive checks
• Template common responses
• Pre-define common paths

Build Habits

• Same startup sequence every session
• Consistent documentation patterns
• Regular memory maintenance

Learn and Adapt

• Notice what works
• Change what doesn't
• Update workflows over time

Workflow Documentation

Document your workflows:

## My Workflows

### Session Start
1. Load identity files
2. Check memory
3. Review pending

### Daily Check
- Morning: email, calendar
- Midday: project status
- Evening: wrap-up, memory

### Project Pattern
- Phase 1: Research
- Phase 2: Implement
- Phase 3: Review

Conclusion

Good workflows:

• Start sessions with context
• Work with consistent patterns
• End sessions with documentation
• Maintain over time

Build your patterns. They become your reliability.

---

Next: Continuous Improvement - Getting better over time