Question 1

Should I use this instead of TypeSafe's own skill?

Accepted Answer

The README flags it as an unofficial community skill that TypeSafe neither made, reviewed nor endorsed, and points at the official skill for that comparison. It does not reproduce the comparison itself in the text captured here, so read both before committing to one.

Question 2

Do I need a TypeSafe API key for it to do anything?

Accepted Answer

No. The skill teaches design patterns on its own. Setting TYPESAFE_API_KEY in the agent's shell adds an optional step where the agent calls the live API to catch wrong field names and questions Jev reads differently than intended, at a stated fraction of a cent per call.

Question 3

Will it work in the agent I already use?

Accepted Answer

There are explicit steps for Claude Code, Claude Desktop, Cowork, Codex, Antigravity CLI, Muse and Muse Code, plus a manual path for any agent that reads SKILL.md: copy the whole skill folder, since SKILL.md links to files beside it. Several paths need a new thread or session before the skill loads.

Question 4

How do I judge whether the advice is any good?

Accepted Answer

By the pattern sketches and the project index, which links 150-plus projects sorted by shape rather than name. The README does not say how those links were verified or how often they are rechecked, so treat the index as a starting point and read the linked repositories before adopting a pattern.

Question 5

How do I store the key so agent shells can actually see it?

Accepted Answer

Keep it in its own file at ~/.config/typesafe/env, written with umask 077, and source it where non-interactive shells will run it: from ~/.zshenv on macOS and Linux zsh, and from the top of ~/.bashrc on Linux bash, because Ubuntu, Debian, Mint and Arch stop reading below their no-terminal guard. Verify with echo ${#TYPESAFE_API_KEY}, which prints the length rather than the key.

Question 6

How fast is a call, and what are the three primitives?

Accepted Answer

Usually 100 to 200 ms. Choice picks one option from a list, such as routing a ticket to billing, shipping or support. Score places the content on a scale you describe, such as grading a pull request from ignores the spec to meets the spec. Noul gives the probability that a yes/no statement is true, such as whether a shell command deletes files outside the project.

Question 7

How does installation work outside Claude Code and Codex?

Accepted Answer

Claude Desktop and claude.ai go through Customize, then Plugins, Add marketplace, Add from repository with Sync automatically left on; it then appears in the desktop app and Cowork on the same account. Antigravity CLI uses agy plugin install with a check through agy plugin list. Muse is set up by pasting a curl-to-bash command into a Muse chat, which copies the folder into ~/workspace/skills and rewrites the SKILL.md header. Muse Code uses muse skills install with --scope user.

Building With TypeSafe Jev

About this project

Highlights

Questions about this entry

More Jev resources like this

From the Jev community