qiushi-skill: a methodology-driven skill set for AI agents to improve strategic reasoning and problem-solving

What it solves

AI Agents often struggle with complex problem-solving, often jumping to conclusions without sufficient investigation or failing to maintain focus on the primary goal. This project provides a structured set of cognitive tools—derived from dialectical materialism and practical philosophy—to help AI agents think more systematically, avoid dogmatism, and execute tasks with greater strategic depth.

How it works

The project implements these philosophies as a collection of "skills" (prompts and guidelines) that can be injected into AI agents. It uses a hierarchical structure:

• General Principle: "Seeking Truth from Facts" (实事求是) acts as the overarching constraint for all judgments.
• Philosophical Foundation: Tools for contradiction analysis and practical cognition to analyze the root of problems.
• Working Methods: Frameworks for investigation, mass-line feedback, and self-criticism to refine results.
• Strategic Tactics: Guidance on protracted strategy, concentrating forces, and overall planning for long-term execution.

These skills are delivered via a CLI tool (npx qiushi-skill) that installs them into various AI environments like Claude Code, Cursor, and others.

Who it’s for

Developers and users of AI agents (such as Claude Code or Cursor) who want to improve their agent's reasoning, strategic planning, and problem-solving capabilities by applying structured methodology.

Highlights

• Nine Cognitive Weapons: A comprehensive set of tools including Contradiction Analysis, Investigation-First, and Protracted Strategy.
• Multi-Platform Support: Easy installation for Claude Code, Cursor, Codex, OpenCode, and more.
• ** uma Workflow Orchestration**: Includes a /workflows layer to chain multiple skills together for complex tasks.
• Grounded in Original Texts: Each skill is backed by original citations to ensure the methodology is faithful to the source material.
• Subagent Prompts: Provides specialized prompts for creating dedicated agents for investigation, contradiction mapping, and feedback synthesis.