DEEP THINKING MODE

Current Time: !date

This task requires extended reasoning. Take time to think through all aspects systematically before acting.

Thinking Protocol

Step 1: Problem Decomposition

Before solving, fully understand:

• •What is the core problem?
• •What are the constraints?
• •What would success look like?
• •What are the risks of getting it wrong?

Step 2: Information Gathering

Gather necessary context:

background_task(agent="explore", prompt="Find relevant code for...")
background_task(agent="librarian", prompt="Research approaches for...")

Read key files directly to build mental model.

Step 3: Alternative Analysis

Consider multiple approaches:

For each, trace through:

• •How would it work step by step?
• •What could go wrong?
• •How would we recover from failure?

Step 4: Edge Case Exploration

Systematically consider:

• •What if input is empty/null/invalid?
• •What if operation times out?
• •What if dependency fails?
• •What about concurrent access?
• •What about scale (10x, 100x, 1000x)?

Step 5: Decision Framework

Choose approach based on:

1. •Correctness — Does it solve the actual problem?
2. •Simplicity — Is it the simplest solution that works?
3. •Robustness — How does it handle failures?
4. •Maintainability — Can others understand and modify it?
5. •Reversibility — How hard to change if wrong?

Step 6: Verification Strategy

Before implementing, plan verification:

• •How will we know it works?
• •What tests should exist?
• •What manual verification is needed?
• •What could we monitor in production?

Output

Write to Obsidian via obsidian_append_content at: $OBSIDIAN_PATH/Thinking/YYYY-MM-DD-topic.md

Note: $OBSIDIAN_PATH must be a vault-relative path (e.g., Projects/myapp), set per-project via direnv. The obsidian_append_content tool expects paths relative to the vault root.

Document Structure

Use this template for the Obsidian document:

@~/.config/opencode/templates/thinking-session.md

Thinking Principles

1. •Understand before solving — Rushing to solution often misses the real problem
2. •Consider alternatives — First idea isn't always best
3. •Trace through logic — Walk through step by step
4. •Verify assumptions — Question what you think you know
5. •Think about failure — Happy path is easy; failure handling is hard
6. •Consider maintenance — Someone else will read this code
7. •Simplify — Can you remove anything?

Think Deeply About

$ARGUMENTS