Why Skill Quality Differs So Dramatically Between People
Writing a Skill requires planning first. Skill is just a tool—the designer is still you.
When using Skills to build a feature, others’ implementations may have far superior stability, speed, and Token consumption—potentially orders of magnitude better than yours.
Take daily digest scraping as an example: you need to fetch articles from 20 data sources, use AI to filter out unwanted content, and aggregate everything into a daily digest.
If you directly ask AI to write the Skill, it will most likely produce a headless browser script that is slow, resource-intensive, and time-consuming.
However, if you research the technical approach with AI before writing the Skill, both reliability and stability will improve dramatically.
Still using the daily digest scraping as our example: after discussing requirements with AI, the plan it provides might look like this:
1. Technology First
For scraping tools: Python scraper + headless browser.
Python is the most widely used language in the web scraping field, which means its ecosystem has accumulated a large number of excellent scraping libraries.
These libraries can help you with: concurrent scraping, automatic retries, and stripping all webpage structure to keep only the main content.
With this approach, running the scraper against any website will directly give you the webpage content, reducing Token consumption by 90% compared to using AI directly.
Of course, this approach won’t cover all cases—some websites have anti-scraping measures. As a fallback, use a headless browser to scrape the site, then have AI summarize the webpage content.
2. Leveraging Built-in Protocols
Many content and news websites provide RSS feeds and Sitemap files.
RSS is designed for feed readers to automatically fetch updated content from websites.
Sitemaps help search engines with SEO indexing.
If any website you want to scrape has either of these, your difficulty in accessing their content drops by 99%.
Because they directly provide all article URLs in the protocol—all you need is a scraper to fetch them.
3. Write the Skill Last
Once you submit this plan to AI, it will generate a fast, compatible, and robust daily digest scraping program for you.
This approach offloads 90% of the AI work to programmatic automation, making your Skill lightning-fast with drastically reduced Token consumption.
This outperforms AI writing code directly by 100x, because the technical solutions used here are proven best practices—many website blocks are automatically handled by the scraping libraries.
The Skill then calls this program directly, and after receiving results, AI filters them according to your predefined preferences to deliver high-quality output.
This is why many experts on social media say AI amplifies the gap between people by 10x or even 100x.
Note: This post is for learning and exchange purposes only. When using scraping technologies, always check websites’ terms of service and obtain consent before scraping content.
