Audit: standup summary blockers + brevity

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.

// Audit: standup summary blockers + brevity — Jev judge eval (typesafe-ai/jev).
// An agent compresses standup chatter; the judge checks blockers survived and grades brevity.
// 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: "Thread: 40 messages about the auth outage, lunch plans, and a deploy. Ana is blocked waiting on DevOps for a Redis failover; Ben will pair with her after lunch.\nSummary: Auth outage contained. Blocker: Ana needs the Redis failover from DevOps (Ben pairing after lunch). Deploy proceeds at 15:00.",
  questions: {
    "blocker_kept": {
      "type": "boolean",
      "instructions": "True only if the summary preserves the blocker, who is blocked, and who unblocks them."
    },
    "brevity": {
      "type": "score",
      "instructions": "Score the summary on signal vs noise. Higher is better.",
      "criteria": [
        "Keeps the noise (lunch) or drops the signal.",
        "Keeps signal but nearly as long as the thread.",
        "Concise with one leftover filler detail.",
        "Three lines: status, blocker with owners, next event."
      ]
    }
  },
});

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