Qwen Code 0.6 Adds Experimental Skills Feature to Rival Claude Code

In the AI programming space, Claude Code has been regarded as a benchmark because of its powerful Agentic workflow and ability to handle complex tasks, and its recent Skills feature has further strengthened its position.

The open‑source Qwen Code project has released version 0.6.0, adding the highly anticipated “Skills” capability as an experimental feature.

Experimental Feature Flag

The Skills functionality is currently experimental and disabled by default to preserve stability. Developers who want to try it must start Qwen with the --experimental-skills parameter, for example:

qwen --experimental-skills

Flexible Loading Mechanism

When enabled, Qwen Code automatically scans the project directory or the user’s configuration directory for Skill definitions. This allows developers to create project‑specific AI skills or maintain a shared skill library—such as a git‑commit assistant or a code‑review tool—that can be reused across projects.

Design Aligned with Claude Code

The Skills system in Qwen Code clearly draws from Claude Code’s successful design. According to PR #1314, the feature was added to support a more complete Agentic workflow. By registering a new skill tool, the core logic can dynamically load and invoke these custom capabilities.

Getting Started

Existing Qwen Code users can upgrade immediately with: npm install -g @qwen-code/qwen-code@latest After logging in via Qwen OAuth, each user receives a daily quota of 2,000 free requests.