think-cog

What This Skill Does

Think-cog transforms your OpenClaw agent into an active thinking partner rather than just a task-executor. Unlike standard skills that rely on a prompt-and-output model, think-cog is built on the iterative cycle of Think → Do → Review → Repeat. It acts as a specialized 'Alfred' that doesn't just provide advice but engages in high-level intellectual exploration, reasoning through complex tradeoffs, and assisting in the emergence of solutions for ambiguous problems. It is designed to bridge the gap between initial ideation and final execution across modalities including research, documentation, data analysis, and prototyping.

Installation

To integrate think-cog into your environment, use the OpenClaw hub tool. Note that this skill requires the primary cellcog skill as a foundation for SDK communication.

clawhub install openclaw/skills/skills/nitishgargiitd/think-cog

Ensure that you have completed the cellcog installation and configuration steps first to avoid dependency errors. Verify your setup with clawhub list after installation.

Use Cases

Think-cog excels where the path to success is not linear. Use it for:

• Architectural Tradeoffs: Evaluating technical stacks (e.g., WebSockets vs. Webhooks) based on team size and infrastructure constraints.
• Business Strategy: Weighing growth models such as enterprise-focused sales versus high-volume self-serve models using real metric analysis.
• Complex Problem Solving: Decomposing vague, multi-faceted problems into actionable steps through iterative conversation.
• Creative Brainstorming: Developing concepts where the solution must be iterated upon rather than generated in a single pass.

Example Prompts

1. "I am struggling to decide between a monolithic or microservices architecture for our new MVP. My team is only two developers, but we expect rapid scaling. Let's walk through the pros and cons given our constraints."
2. "I need to refine our product roadmap. We are currently split between adding a mobile app or deepening our desktop analytics suite. Analyze these options considering our current 15% churn rate."
3. "Help me think through the design of a new data schema for a user-preference engine that needs to handle high-cardinality attributes without hurting query performance."

Tips & Limitations

• Be Conversational: Think-cog performs best when you treat it as an collaborator. Do not expect an immediate 'done' result; expect a dialogue that leads to a decision.
• Context Loading: Provide raw data or specific metrics when possible. The quality of the 'thinking' is directly proportional to the context you provide.
• Iterate Often: Use the feedback loop. After the agent presents an idea, challenge the assumptions or provide new constraints to narrow the focus.
• Limitations: This skill is not intended for simple retrieval tasks. If you already know the solution and just need a command run, use standard atomic skills instead.