jev-agent-setup

Installs the jev CLI and managed Jev instruction blocks into global agent configuration files.

1|Updated May 28, 2026
One-click install
npx skills add https://github.com/dhanesh/agent-skills --skill jev-agent-setup-dhanesh
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: jev-agent-setup
Source: https://github.com/dhanesh/agent-skills/tree/main/jev-agent-setup
Command: npx skills add https://github.com/dhanesh/agent-skills --skill jev-agent-setup-dhanesh

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes assets (resource) and references (resource) components.

What problem does it solve? Coding agents like Claude Code, Codex, and Gemini CLI each keep their own global instruction files, so making every agent on a machine offload rank/classify/yes-no decisions to TypeSafe's Jev requires editing four separate files by hand and keeping them in sync. This Skill installs the jev CLI and a managed BCP 14 instruction block into all of them idempotently, with drift detection and safe uninstall. ## Core Features & Use Cases - Multi-agent installation: Writes a managed Jev block into ~/.claude/CLAUDE.md, ~/.agents/AGENTS.md, $CODEX_HOME/AGENTS.md, and ~/.gemini/GEMINI.md, preserving all content outside the markers and backing up files before first modification. - Block or mirror modes: --mode block adds only the Jev rules; --mode mirror gives non-Claude agents a managed copy of the whole CLAUDE.md with @imports inlined. - Drift checking and verification: --check reports OK/STALE/MISSING per target with exit 1 on drift, refuses malformed markers with exit 3 before writing, and supports a smoke-test Noul call plus jev log review of per-project call logs. - Use Case: You set up a new laptop and want Claude Code, Codex, and Gemini CLI to all route their System One decisions through Jev. Run the installer once, add your TypeSafe API key to ~/.config/typesafe/env, and verify with --check. ## Quick Start Ask your agent to set up Jev for all coding agents on this machine and verify the installation with a drift check.

Frequently Asked Questions about jev-agent-setup

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make Claude Code, Codex, and Gemini CLI all use Jev?▼

Run the install_jev_setup.py installer, which writes a managed Jev instruction block into each agent's global instruction file and installs the jev CLI into ~/.local/bin. Re-running is idempotent, and --check verifies all targets afterward.

What is the difference between jev-agent-setup and the typesafe-ai skill?▼

typesafe-ai teaches an agent how to design Jev judgments inside an application being built. jev-agent-setup makes every agent on the machine route its own working-cycle decisions through Jev, persistently, via global instruction files and the jev CLI.

Does the installer overwrite my existing CLAUDE.md or AGENTS.md content?▼

No. Content outside the BEGIN/END jev-agent-setup markers is never touched, and the first modification backs the file up to a .bak-jev-agent-setup copy. Uninstall restores pre-existing files byte for byte.

Why does the Jev installer refuse to run with exit code 3?▼

Exit 3 means a target file has malformed markers, such as a BEGIN without an END or duplicates. The installer validates every target before writing anything, so you must fix the markers by hand and re-run.

What do jev CLI exit codes 2, 3, and 4 mean?▼

Exit 2 means bad input or a missing API key, exit 3 means Jev is unavailable (timeout, 429, or 5xx) and may be retried once, and exit 4 means TypeSafe rejected the request (400/401/403/404/422) and it must not be retried unchanged.

Where is the TypeSafe API key stored and is it ever written by the installer?▼

The key lives in TYPESAFE_API_KEY with fallback to ~/.config/typesafe/env, which the installer creates only as a 0600 placeholder if missing. The installer never prints or writes a real key; you edit the file yourself.