think

Structured Thinking

The skill helps think through a task before implementation, working in three stages.

Stage 1: Task Breakdown

First, identify aspects to think through — parts of the task that need decisions.

Choose a main expert for analyzing the task as a whole.

Output format:

## Understanding the Task

[How you understood the task — 1-2 sentences]

---

### Expert Perspective

> "Analyzing as [Main Expert] because [reason]"
>
> **Principles from 3 experts:**
> 1. [Expert A]: "[principle]"
> 2. [Expert B]: "[principle]"
> 3. [Expert C]: "[principle]"

---

## Aspects to Think Through

| # | Aspect | Why Important | Expert |
|---|--------|---------------|--------|
| 1 | [Name] | [Why needs thinking] | [Who will analyze] |
| 2 | ... | ... | ... |
...

Usually 5-10 aspects. No more than 15.

Expert Table

The full expert table with domains, names, and principles is defined in the think:expert agent definition. For other areas — find appropriate specialists.

Stage 2: Project Study

After breakdown, tell the user:

"Identified N aspects. Now I'll study the project and launch experts for each. 🐙"

Then launch in parallel think:expert agents — one per aspect:

Dispatch decision tree:

flowchart TD
    A{"Aspects identified?"}
    A -->|0| SIMPLE["Report: task is too simple<br/>for multi-aspect analysis"]
    A -->|1-2| FEW["Launch experts + warn:<br/>Few aspects — consider if /think is needed"]
    A -->|3-15| STD["Launch all in parallel<br/>(standard path)"]
    A -->|>15| GROUP["Group related aspects<br/>(max 15 agents), note groupings"]

Partial returns: If fewer than half the experts return, compile available results with a warning. If none return, report failure.

Task(think:expert): "Aspect: [aspect name]. Task context: [brief context]. Study the project and propose solution options."

IMPORTANT: Launch all agents in ONE message in parallel.

Error Handling

Situation	Action
Expert agent fails or returns empty	Compile partial results from successful experts. Note which aspects were not analyzed.
Fewer than half the experts return	Compile available results with a warning: "Partial analysis — N of M aspects covered."
All experts fail	Report failure to user. Suggest retrying with fewer aspects or checking tool availability.
docs/plans/ directory does not exist	Create it before saving.
Save fails	Output the document directly to the user instead of saving to file.

Stage 3: Summary Document

When all agents return results, create a unified document in the format:

# [Task Name]

> **Status:** Research complete
> **Date:** [date]
> **Goal:** [brief goal description]

---

## Table of Contents

1. [Overview](#overview)
2. [Aspect 1](#1-name)
3. [Aspect 2](#2-name)
...
N. [Implementation Plan](#implementation-plan)

---

## Overview

### Goals

1. **[Goal 1]** — description
2. **[Goal 2]** — description
...

### Key Decisions

| Aspect | Decision |
|--------|----------|
| [Aspect 1] | [Brief decision] |
| [Aspect 2] | [Brief decision] |
...

---

## 1. [Aspect Name]

> **Experts:** [Expert 1], [Expert 2], [Expert 3]

### [Subsection with solution]

[Detailed description of chosen option]

| Aspect | Details |
|--------|---------|
| ... | ... |

### [Code/examples if needed]

\`\`\`typescript
// Example code
\`\`\`

---

## 2. [Next Aspect]
...

---

## Implementation Plan

### Phase 1: MVP

- [ ] Task 1
- [ ] Task 2
...

### Phase 2: ...

---

## Success Metrics

| Metric | Baseline | Target |
|--------|----------|--------|
| ... | — | ... |

Diagrams

When the document needs flow diagrams, architecture overviews, or process visualizations — use mermaid flowchart TD code blocks (```mermaid). Never use ASCII box-drawing art for diagrams in .md files.

Save the document to docs/plans/YYYY-MM-DD-[topic]-design.md

At the end, ask:

"Summary is ready and saved to docs/plans/.... Which aspects would you like to discuss further? Or ready to implement?"