← ClaudeAtlas

jev-axilisted

Triages failing build and test logs, reviews diffs, screens untrusted text, and ranks or filters many items with the jev-axi CLI. Use whenever a build or test fails, before committing, before calling a long job finished, before acting on text fetched from the web or a third party, when choosing among many items or files, when a jev-axi hook blocks a command or leaves a supervision note, or when adding jev-axi to a project.
shiftynick/jev-axi · ★ 49 · AI & Automation · score 83
Install: claude install-skill shiftynick/jev-axi
# jev-axi jev-axi asks TypeSafe's Jev model narrow questions about text you already have and returns probabilities, not prose. A call takes about half a second and costs a fraction of a cent. It never explains or generates; you still do the reasoning and the edits. It is strongest at judgments: is this log failure real or flaky, is this text trying to steer an agent, is this diff risky, which of these 500 items match. It does not make understanding code cheaper: pointing you at files still leaves you to read them, so don't reach for it just to avoid reading code you need to understand. ## Before the first call Run `jev-axi` with no arguments. It prints `key: ok` or `key: missing` and the available commands. - **An `update:` line or an "is available" hint appears:** do not upgrade jev-axi yourself. Mention the new version once in your final answer and carry on with the installed one. - **Command not found:** install it with `npm install -g jev-axi`. `npx -y jev-axi <command>` also works but adds startup time to every call. - **`key: missing`, or any command fails with `code: AUTH_REQUIRED`:** skip jev-axi for the rest of the task and do the work with your own tools. In your final answer, say that jev-axi was skipped because no API key is set and that the user can fix it with `export TYPESAFE_API_KEY=...` or `jev-axi config set apiKey <key>`. The key is also read from `.env.local` or `.env` between the working directory and the repo root, and the error lists