wtf is skill-authoring?

Skill Authoring is a skill file meant to teach AI coding agents how to write better skills of their own. A skill, in this context, is a packaged set of instructions that an agent reads to handle a particular kind of task. The project points out that when you ask an agent to create one of these skill files, the result often looks professional but does not actually work well: descriptions that never get triggered at the right moment, bodies stuffed with obvious explanations that waste the agent's available context, and folder layouts copied blindly from ordinary software projects rather than designed for how agents actually read them. This project tries to fix that by teaching agents to tell a usable skill apart from a good one, and it uses its own structure as a working example of the advice it gives. It opens with a routing table so a reader can jump straight to the part they need, and it pairs each of its design principles with a short check question and a warning about the common way that principle gets misapplied, keeping the rule, the test, and the failure pattern together in one place. Installing it means copying a single file called SKILL.md into the skills folder used by your agent, whether that is OpenCode, Codex, Claude Code, or Cursor, each of which has its own expected folder location. The whole skill fits in 328 lines and includes a quickstart with a minimal copy paste skeleton, six design principles with built in anti pattern warnings, a decision tree for choosing between creating a skill, saving a reference document, or just writing a one time prompt, four common patterns with concrete examples, and a basic quality checklist for self review. The author notes this project is meant to complement, not replace, skills focused on prompt writing itself, since this one is about structure and file placement rather than exact wording. It draws on official documentation from Anthropic and OpenAI along with community writing on the same topic, and it is released under the MIT license.