reasoning-patterns

name	reasoning-patterns
version	1.0.0
category	software-development
description	Select and apply reasoning patterns for any task — Plan-Execute-Verify, ReAct, Reflexion, or Tree of Thoughts. Loaded at session start to choose the right thinking strategy.
pinned	true

Reasoning Patterns — Choose How to Think

Loaded at session start. After classifying the task with agent-flow, choose the reasoning pattern that fits.

The Patterns

Pattern	When to Use	Structure
Plan-Execute-Verify	Default — most tasks	Write plan → execute steps → verify each step with tool output
ReAct	Debugging, exploration	Reason about the situation → Act (one tool call) → Observe result → Repeat
Reflexion	Quality-critical work (add to any pattern)	Execute → Self-critique → Fix → Re-verify before delivering
Tree of Thoughts	Design decisions with trade-offs	Generate 3+ approaches → Evaluate each → Select best → Implement

How to Choose

1. Start with Plan-Execute-Verify unless the task clearly demands another pattern
2. Switch to ReAct when something unexpected happens, you're debugging, or exploration is needed
3. Add Reflexion on top of any pattern when the output quality is critical (production code, security, user-facing docs)
4. Use Tree of Thoughts for architectural decisions, comparing approaches, or any task where the first idea might not be the best

State Your Choice

Always state which pattern you're using:

"Using Plan-Execute-Verify with Reflexion check."

"Switching to ReAct — need to investigate this failure first."

Why This Exists

Without an explicit reasoning pattern, agents default to whatever feels natural — which may not match the task's needs. Debugging needs ReAct, not a rigid plan. Design needs ToT, not a single pass. This skill makes the choice deliberate and auditable.