Skills package instructions and sometimes scripts that an agent can use for a specific task. They can save you from explaining the same workflow every time. Community packages still need review: instructions and code can access data or perform actions within the agent’s permissions.
1. Self-Improvement
The Self-Improvement skill records errors, corrections, and lessons in a .learnings/ directory for use in later work. This is stored guidance, not an automatic upgrade to the underlying model.
2. Multi Search Engine
Multi Search Engine provides instructions for several search engines. The original listing described 16 engines. Having alternatives can help with different languages and sources, but results still need checking.
3. Humanizer
Humanizer offers editing guidance intended to reduce repetitive, exaggerated, and formulaic prose. It is useful when writing with an agent, though a style prompt does not establish authorship or guarantee any detector result.
4. Free Ride
Free Ride helps select available free-tier models on OpenRouter. Free models can have request limits, token limits, or changing availability; “free” does not mean unlimited. The package needs an API key, so review how it stores and uses that credential.
5. Stock Analysis
Stock Analysis uses Yahoo Finance data to organize company metrics, dividends, and market information. It can help structure research, but delayed data, incomplete coverage, and model errors limit what you can conclude from its output.
I also wrote about a chat-history character skill (Chinese original) and a character companion for coding (Chinese original). Browse a package’s files and permissions before installation, and choose skills for tasks you actually need. The OpenClaw skill documentation explains the loading mechanism.
Adapted from the original Chinese article, published on April 20, 2026.
Comments NOTHING