serverinoGive it the reference, not your memory of it
Serverino is small — a handful of attributes and two structs — but it is also small enough to be rare in a model's training data, so a model left to guess will fill the gaps with vibe.d or Node idioms. Hand it these files instead. They work with any tool: there is nothing vendor-specific about a text file.
Three files, served from here
| File | What it is |
|---|---|
/llms.txt |
A page of overview, in the llms.txt shape. Cheap to paste anywhere. |
/llms-full.txt |
The whole API in one file: configuration, every attribute, Request,
Output, WebSockets, worked examples. Around 700 lines. |
/SKILL.md |
The rules that are easiest to get wrong, ready to install as a skill.
/AGENTS.md is the very same text without
the front matter, for tools that want a plain rules file to save as
AGENTS.md, CLAUDE.md or .cursorrules. |
Ask your agent to install it
Anything that can read a URL can do this for you, so there is nothing to download by hand. Paste this and let it work:
Install the skill at https://serverino.dev/SKILL.md. It is the reference for serverino, the D web server library I am using.
Claude Code drops it in ~/.claude/skills/; other tools save it as their rules
file. Either way the model stops guessing, and what it reads is the same text this page
links above.