workflow-orchestration

WorkflowOrchestration - Development Excellence Methodology

Auto-loaded for development work. This skill enforces rigorous development methodology that eliminates common failure modes and maximizes code quality.

Core Philosophy

High-quality software comes from disciplined process, not heroic effort. This methodology prevents the most common development failures:

• Starting implementation before understanding the problem (Plan Mode Default)
• Thrashing in a broken state instead of re-planning (Stop & Re-plan)
• Working in a cluttered context window (Subagent Strategy)
• Repeating the same mistakes (Self-Improvement Loop)
• Claiming completion without proof (Verification Before Done)
• Accepting hacky solutions when elegant ones exist (Demand Elegance)
• Waiting for hand-holding on obvious fixes (Autonomous Bug Fixing)

---

1. Plan Mode Default

CRITICAL: Enter plan mode for ANY non-trivial task (3+ steps or architectural decisions)

When to Plan

Enter plan mode when ANY of these are true:

• Task requires 3+ distinct steps
• Architectural decisions need to be made
• Multiple files/components will be affected
• User says "plan first" or requests a plan
• You're not 100% certain about the approach
• Something went sideways (stop immediately, don't push forward)

Planning Process

# 1. Write detailed plan to task file
# Location: ~/Nextcloud/PAI/tasks/todo.md

# 2. Plan structure (required sections):
# - OBJECTIVE: What are we building/fixing?
# - CONTEXT: What's the current state?
# - CONSTRAINTS: What limitations exist?
# - APPROACH: High-level strategy
# - STEPS: Numbered checklist items [ ]
# - VERIFICATION: How will we prove it works?
# - ROLLBACK: How to undo if things break?

# 3. Check in with user BEFORE starting implementation
# Present the plan, wait for approval

# 4. Track progress as you work
# Mark items complete: [x] in todo.md

# 5. Explain changes at each step
# Add high-level summary after each major step

# 6. Document results
# Add REVIEW section to todo.md with:
# - What worked
# - What didn't
# - Lessons learned
# - Final verification results