@leeoxiang: I realized that writing skills also requires considering compatibility. For example, implementing a capability like AskUserQuestion requires writing adapters for Claude Code and Codex separately. Isn't this just like the old days of frontend development when we had to adapt for IE6?
Summary
The author points out that when writing AI agent skills, one must consider compatibility across different platforms (such as Claude Code and Codex), similar to how frontend development needed to adapt for different browsers in the past.
Similar Articles
@shao__meng: The Internal Design, Iteration, and Maintenance of Agent Skills at Perplexity. The public version of Perplexity Agents' internal standards presents a counter-intuitive core argument: writing a Skill is not about writing code, but about building context for the model. Applying the instinct of engineers writing code directly to Skills...
The Perplexity team has published guidelines for the design, iteration, and maintenance of Agent Skills, emphasizing that writing Skills is not traditional coding but rather constructing context for the model. The article proposes a counter-intuitive methodology focused on evaluation-first approaches, progressive loading, and optimizing Agent behavior by handling edge cases (Gotchas).
@yetone: Since this article is so great, I turned it into an Agent Skill. You can install this Skill using your own Coding Agent, so you can easily refactor or develop a cross-platform desktop app with near-native performance using 'best practices'…
A GitHub repository packaging an Agent Skill that distills best practices for building cross-platform desktop apps with near-native performance, inspired by Raycast 2.0's architecture.
@AYi_AInotes: Say a hot take: In the AI era, the most valuable skill is no longer writing code. Being able to explain code clearly will become increasingly important! Becoming increasingly important! @trq212, a senior engineer on the Anthropic Claude Code team, took less than two years to make his technical articles reach stable...
This article explores the importance of technical writing in the AI era, citing the case of Anthropic employee @trq212 who achieved millions of page views through his 'plant first, harvest later' writing methodology, emphasizing the value of sharing real experiences and maintaining a personal voice.
@shao__meng: Why do Claude Code, Cursor, Codex, Aider, and Cline exhibit different agent behaviors despite potentially sharing the same underlying models? @addyosmani argues: It's due to the "shell" above the model — the Harness, which includes "prompts, ...
The article discusses how Addy Osmani argues that the performance difference between AI coding agents like Claude Code, Cursor, and Cline stems from their 'Harness'—the layer of prompts, tools, and constraints around the model—rather than the underlying model itself. It details best practices for harness engineering, including hooks, sandboxing, and context management, to bridge the gap between model capability and actual agent performance.
@yihui_indie: Asking the community two questions: 1. How do you manage your growing collection of SKILLs? 2. How do you sync SKILLs between Claude Code and Codex? Are there any good solutions for this?
The user is asking for advice on managing and syncing skills (SKILLs) for Claude Code and Codex.