Check: beginner fitness plan safety + progression

The job

Jev judge eval (1 boolean + 1 score) over a fixed state : runnable via experimental_evaluate.

The prompt

Copy this into Muse and replace anything in [square brackets] with your own details.

// Check: beginner fitness plan safety + progression — Jev judge eval (typesafe-ai/jev).
// An agent writes a couch-to-5K plan; the judge grades it and checks rest days.
// Model: typesafe-ai/jev via the Vercel AI Gateway ($0.042/1M input tokens).
// Auth: set AI_GATEWAY_API_KEY (vck_…) in the environment — never paste the key here.
// For zero-data-retention / no-training workloads, add
// providerOptions: { gateway: { … } } to the evaluate call below.

import { experimental_evaluate } from 'ai';

const result = await experimental_evaluate({
  model: 'typesafe-ai/jev',
  state: "Plan (8 weeks): 3 runs/week, walk-run intervals progressing 60s→5min runs. Rest days between every session; week 4 deload at half volume. Stop rules: sharp pain = stop and see a clinician; includes 5-min warm-up/cool-down each session.",
  questions: {
    "plan_quality": {
      "type": "score",
      "instructions": "Score this beginner running plan. Higher is better.",
      "criteria": [
        "Dangerous (no progression, no rest, or ignores pain).",
        "Directionally fine but missing progression, rest, or stop rules.",
        "Progressive with rest, but no deload or stop rules.",
        "Progressive, rest days, deload week, warm-ups, and stop rules."
      ]
    },
    "rest_scheduled": {
      "type": "boolean",
      "instructions": "True only if rest or recovery days are explicitly scheduled between sessions."
    }
  },
});

console.log(JSON.stringify(result, null, 2));