building-with-typesafe-jev

<p align="center"> <strong>Building with TypeSafe Jev</strong><br> <em>Typed decisions with calibrated confidence, for your coding agent.</em><br> <em>Links to 150+ community projects, sorted by shape, with a code sketch for each.</em> </p> <p align="center"> <a href="LICENSE"><img src="https://img.shields.io/github/license/aaddrick/building-with-typesafe-jev?style=flat" alt="License"></a> <a href=".github/workflows/checks.yml"><img src="https://img.shields.io/github/actions/workflow/status/aaddrick/building-with-typesafe-jev/checks.yml?label=checks&style=flat" alt="Checks"></a> </p> <p align="center"> <a href="https://www.linkedin.com/in/aaddrick/">Connect on LinkedIn!</a> </p> <p align="center"> <strong>English</strong> · <a href=".github/readme/README.zh-CN.md">简体中文</a> · <a href=".github/readme/README.ja.md">日本語</a> · <a href=".github/readme/README.ko.md">한국어</a> · <a href=".github/readme/README.vi.md">Tiếng Việt</a> · <a href=".github/readme/README.pt-BR.md">Português (BR)</a> · <a href=".github/readme/README.it.md">Italiano</a> </p>

[!NOTE] This is an unofficial, community skill. It is not made, reviewed, or endorsed by TypeSafe AI. TypeSafe publishes its own skill at typesafe-ai/skills. See How this differs from the official skill.

Coding agents treat Jev like one more chat model. This skill teaches them to design for it: typed questions, calibrated confidence, and links to 150+ community projects, sorted by how they work, with a code sketch for each pattern. It installs in Claude Code, Codex, Antigravity CLI, Muse, and Muse Code.

Jev is a System One model. It does not write text. You send it content and a set of typed questions, and it answers each one with a value and a calibrated probability, usually in 100 to 200 ms:

• Choice picks one option from a list. Example: route a ticket to billing, shipping, or support.
• Score places the content on a scale you describe. Example: grade a pull request from "ignores the spec" to "meets the spec".
• Noul gives the probability that a yes/no statement is true. Example: "this shell command deletes files outside the project."

Install

<details> <summary><strong>Claude Code</strong></summary>

claude plugin marketplace add aaddrick/building-with-typesafe-jev

claude plugin install building-with-typesafe-jev@building-with-typesafe-jev

The skill loads on its own when you work on Jev code. To load it by hand, type:

/building-with-typesafe-jev:building-with-typesafe-jev

</details> <details> <summary><strong>Codex</strong></summary>

codex plugin marketplace add aaddrick/building-with-typesafe-jev

codex plugin add building-with-typesafe-jev@building-with-typesafe-jev

Start a new thread. Codex loads the skill when the task matches. To load it by hand, type:

$building-with-typesafe-jev:building-with-typesafe-jev

</details> <details> <summary><strong>Antigravity CLI</strong></summary>

agy plugin install https://github.com/aaddrick/building-with-typesafe-jev

Check that it installed:

agy plugin list

Start a new session. Antigravity CLI loads the skill when the task matches. To load it by hand, type:

/building-with-typesafe-jev:building-with-typesafe-jev

Coming from Gemini CLI? If agy plugin import gemini brought this extension over, run the install command above anyway so the current copy replaces the imported one.

</details> <details> <summary><strong>Muse (muse.ai)</strong></summary>

Muse loads skills from ~/workspace/skills/ on its own computer. Paste this command into a Muse chat and ask Muse to run it:

curl -fsSL https://raw.githubusercontent.com/aaddrick/building-with-typesafe-jev/main/scripts/install_muse.sh | bash

The script copies the skill folder there and rewrites the SKILL.md header into the shape Muse reads. Start a new chat. Muse loads the skill when the task matches. To update, run the command again.

</details> <details> <summary><strong>Muse Code</strong></summary>

Clone the repository:

git clone https://github.com/aaddrick/building-with-typesafe-jev.git

Install the skill for every project:

muse skills install building-with-typesafe-jev/skills/building-with-typesafe-jev --scope user

Check that it installed:

muse skills list

Start a new session. Muse Code loads the skill when the task matches. To load it by hand, type:

/building-with-typesafe-jev

</details> <details> <summary><strong>Any other agent that reads SKILL.md</strong></summary>

Copy the skills/building-with-typesafe-jev/ folder into your agent's skills folder. Keep the whole folder. SKILL.md links to the files beside it.

</details>

Set up an API key (optional, recommended)

The skill works without a key. With TYPESAFE_API_KEY set in the agent's shell, the agent can check its design against the live API before the code reaches your project. That catches wrong field names, and questions Jev reads differently than you meant. Each call costs a fraction of a cent.

<details> <summary><strong>Create, store, and troubleshoot a key</strong></summary> <br> <details> <summary><strong>Create a key</strong> (four steps in the TypeSafe console)</summary>

Step 1. Sign in at console.typesafe.ai and open API Keys in the sidebar.

<img src=".github/assets/api-key/step-1.png" alt="The TypeSafe console home page. An amber box and arrow point at API Keys in the left sidebar." width="100%">

Step 2. Click Create key at the top right.

<img src=".github/assets/api-key/step-2.png" alt="The API keys page. Existing keys are blurred. An amber box and arrow point at the Create key button at the top right." width="100%">

Step 3. Name the key after where it will live, such as the machine or the agent. Then click Create key.

<img src=".github/assets/api-key/step-3.png" alt="The Create API key dialog with the name my-coding-agent typed in. An amber box and arrow point at the name field and the Create key button." width="100%">

Step 4. Copy the key now. The console shows it once. If you lose it, create a new one and revoke the old one.

<img src=".github/assets/api-key/step-4.png" alt="The API key created dialog. The key value is masked. An amber box and arrow point at the Copy button." width="100%"> </details> <details> <summary><strong>Store the key</strong> (macOS, Linux, Windows)</summary>

Keep the key in its own file, readable only by you, and export it as TYPESAFE_API_KEY. Agents often start shells without a terminal, so each section puts the key where those shells can see it. Pick your system.

<details> <summary><strong>macOS</strong> (zsh, the default shell)</summary>

Save the key to a private file:

mkdir -p ~/.config/typesafe && umask 077 && printf 'export TYPESAFE_API_KEY=%s\n' 'YOUR_KEY' > ~/.config/typesafe/env

Load it from ~/.zshenv. Every zsh reads that file, including shells that agents start without a terminal. ~/.zshrc is read only by interactive shells.

echo '[ -f ~/.config/typesafe/env ] && . ~/.config/typesafe/env' >> ~/.zshenv

Open a new terminal and check it. The command prints the length of the key, not the key:

echo ${#TYPESAFE_API_KEY}

</details> <details> <summary><strong>Linux with bash</strong> (Ubuntu, Debian, Mint, Fedora, Arch)</summary>

Save the key to a private file:

mkdir -p ~/.config/typesafe && umask 077 && printf 'export TYPESAFE_API_KEY=%s\n' 'YOUR_KEY' > ~/.config/typesafe/env

Load it from the top of ~/.bashrc. Ubuntu, Debian, Mint, and Arch start ~/.bashrc with a line that stops early when no terminal is attached. A line below that guard never runs for agent shells. The top of the file is safe on every distribution:

sed -i '1i [ -f ~/.config/typesafe/env ] \&\& . ~/.config/typesafe/env' ~/.bashrc

Open a new terminal and check it:

echo ${#TYPESAFE_API_KEY}

</details> <details> <summary><strong>Linux with zsh</strong></summary>

Follow the macOS steps. zsh reads ~/.zshenv the same way on Linux.

</details> <details> <summary><strong>Linux with fish</strong></summary>

fish reads every file in ~/.config/fish/conf.d/, with or without a terminal:

mkdir -p ~/.config/fish/conf.d; and echo 'set -gx TYPESAFE_API_KEY YOUR_KEY' > ~/.config/fish/conf.d/typesafe.fish; and chmod 600 ~/.config/fish/conf.d/typesafe.fish

string length -- $TYPESAFE_API_KEY

</details> <details> <summary><strong>Windows</strong> (PowerShell)</summary>

Store the key as a user environment variable. New terminals and apps see it. Terminals that are already open do not:

[Environment]::SetEnvironmentVariable('TYPESAFE_API_KEY', 'YOUR_KEY', 'User')

Open a new terminal and check it:

$env:TYPESAFE_API_KEY.Length

WSL: Windows variables do not reach WSL by default. Inside WSL, follow the Linux with bash steps.

</details> <details> <summary><strong>Desktop apps and IDE extensions</strong></summary>

An app you start from the dock, the start menu, or a desktop launcher does not read your shell files.

• Windows: the user environment variable above already covers these apps.
• Linux (systemd): add the line TYPESAFE_API_KEY=YOUR_KEY to ~/.config/environment.d/typesafe.conf, then log out and back in.
• macOS: start the app from a terminal, or set the variable in the app's own settings. macOS has no simple per-user file that desktop apps read.

</details> </details> <details> <summary><strong>If the agent cannot see the key</strong></summary>

Ask the agent to run echo ${#TYPESAFE_API_KEY} (or $env:TYPESAFE_API_KEY.Length on Windows). If it prints 0 or nothing, check these in order:

• You started the agent before you stored the key. Agent shells copy the environment of the program that started them. Quit the agent and start it from a new terminal.
• You started the agent from the dock, the start menu, or an IDE. Those apps do not read your shell files. See "Desktop apps and IDE extensions" above.
• Codex filters the environment. If ~/.codex/config.toml sets include_only under [shell_environment_policy], add TYPESAFE_API_KEY to it. If it sets ignore_default_excludes = false, Codex drops every variable with KEY in its name. Remove that line.
• WSL. Windows variables do not reach WSL. Store the key inside WSL with the Linux steps.

</details> </details>

How it was tested

We gave a coding agent six Jev tasks, such as a support-ticket triage function and an approval gate for shell commands. Each task ran 10 times with this skill, with no skill, and with the official skill, each arm in its own isolated container. The agent had the docs but no API key, so graders checked the code it wrote. Checks that need judgment went to three LLM judges from three providers (Claude Opus, GPT-6 Sol, Kimi K3), and the majority decided.

| | No skill | This skill | Official skill | |---|---:|---:|---:| | Average score | 0.65 ± 0.05 | 0.96 ± 0.02 | 0.77 ± 0.04 |

Score is the share of checks passed, averaged over 10 runs per task.

Where the skill made the difference (runs out of 10 that passed):

| The agent's code... | No skill | This skill | Official skill | |---|---:|---:|---:| | counted in code, instead of asking Jev for a number | 1 | 8 | 0 | | named a field of the input in its questions | 0 | 10 | 1 | | pinned or logged the Jev model version | 0 | 10 | 0 | | gave the department list a catch-all option | 3 | 10 | 6 | | kept more than one path while walking 1,200 categories | 1 | 10 | 7 | | kept a plain-code backstop for destructive commands | 7 | 10 | 2 | | read score as a position from