Posts about Artificial intelligence

What agent skills are, when they are worth it and how a development team should organize them: the open standard, structure, evals, security and real cases.

A skill without evals is an assumption. Prompts that should and should not trigger it, comparing with and without the skill, and what to do when the model changes.

A team's knowledge is spread across docs, GPTs, CLAUDE.md files and people's heads. How to turn it into AI skills without copying dead documentation.

A Claude Code plugin bundles skills, commands and hooks. How to structure one, what each piece does and why critical rules belong in a hook, not a skill.

Creating a skill is easy; getting it to trigger at the right time is not. How to write the description, the router pattern, scripts and SKILL.md mistakes.

Many agent skills look like a good idea and only add complexity. How to tell a process that needs a skill from one that needs a procedure or a script.