think

Route reasoning tasks to thinking modes and emit schema-aligned JSON outputs.

1|Updated Apr 13, 2026

Search Tags:#routing #reasoning #ai-architecture #structured-output #mode-selection #think

Ready to run

npx skills add https://github.com/danielsimonjr/deepthinking-plugin --skill think-danielsimonjr

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: think
Source: https://github.com/danielsimonjr/deepthinking-plugin/tree/main/skills/think
Command: npx skills add https://github.com/danielsimonjr/deepthinking-plugin --skill think-danielsimonjr

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Route reasoning tasks to the appropriate thinking mode, enabling targeted method application and consistent outputs.

Core Features & Use Cases

Parse the user invocation to determine an explicit mode or rely on the mode-index decision tree for auto-selection.
Load the corresponding category skill, apply its method, and produce a structured JSON output aligned with the mode's schema.
Support end-to-end routing across the 34 modes with a single entry point and clear fallbacks.

Quick Start

Invoke /think bayesian "update my belief that the service is down given new evidence" to see Bayesian updating in action.

Frequently Asked Questions about think

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I route reasoning tasks to the appropriate thinking mode automatically?▼

You can route reasoning tasks automatically by relying on the mode-index decision tree to select the appropriate thinking mode, apply its method, and generate a structured JSON output.

How does structured output routing work for AI reasoning architectures?▼

Structured output routing works by parsing the invocation, loading the corresponding category skill, applying its method, and emitting a JSON object aligned with the chosen mode's schema.

Can I explicitly select a specific reasoning mode like Bayesian updating?▼

Yes, you can explicitly select a reasoning mode like Bayesian updating by specifying it in the invocation, which bypasses the auto-recommendation decision tree.

What is the best way to manage multiple reasoning modes from a single entry point?▼

The best way to manage multiple reasoning modes is using a single entry point that supports end-to-end routing across 34 modes with clear fallbacks for consistent structured outputs.

Do I need to specify a reasoning mode if I want structured output?▼

No, you do not need to specify a mode manually; the skill can auto-recommend a mode using its decision tree and still produce a structured JSON output matching the mode's schema.

What happens when a reasoning task does not match a specific mode?▼

When a reasoning task does not match an explicitly requested mode, the system applies an auto-recommendation decision tree to route the task and provides clear fallbacks for structured output.