Recipe catalog / handoff-completeness

Grade handoff completeness

How ready is item, a work description being handed to another agent or person, to be picked up without asking questions, on a five-level rubric?

You need to grade a task description before delegating it, so an orchestrator can enrich it or ask for missing pieces instead of handing off something a worker will bounce back.

Explore this recipe interactively · Source and implementation guide

Use handoff-completeness in TypeScript

Install with npm install jev-recipes. Requires Node.js 22.9 or newer and ES modules. Set TYPESAFE_API_KEY in your server environment for live calls, which send input to TypeSafe and use API quota. See the installation guide.

import { handoffCompleteness } from 'jev-recipes/handoff-completeness';

const result = await handoffCompleteness({
  "item": "Add rate limiting to the public /search endpoint. Use the existing RateLimiter middleware in src/middleware/rate-limit.ts and the Redis instance already configured for sessions. Done when requests beyond 60 per minute per API key return 429 with a Retry-After header and the integration test in tests/api/search.test.ts covers the limit.",
  "minConfidence": 0.8
});
console.log(result);

Input contract

{
  "input": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "item": {
        "type": "string"
      },
      "minConfidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      }
    },
    "required": [
      "item"
    ]
  },
  "result": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "model": {
        "type": "string"
      },
      "usage": {
        "type": "object",
        "properties": {
          "input_tokens": {
            "type": "integer",
            "minimum": 0,
            "maximum": 9007199254740991
          },
          "output_tokens": {
            "type": "integer",
            "minimum": 0,
            "maximum": 9007199254740991
          }
        },
        "required": [
          "input_tokens",
          "output_tokens"
        ],
        "additionalProperties": false
      },
      "status": {
        "type": "string",
        "enum": [
          "ready",
          "review"
        ]
      },
      "score": {
        "type": "number",
        "minimum": 0
      },
      "level": {
        "type": "integer",
        "minimum": 0,
        "maximum": 9007199254740991
      },
      "confidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      },
      "probabilities": {
        "type": "object",
        "propertyNames": {
          "type": "string"
        },
        "additionalProperties": {
          "type": "number",
          "minimum": 0,
          "maximum": 1
        }
      },
      "completeness": {
        "type": "string",
        "enum": [
          "bare",
          "goal",
          "inputs",
          "defined",
          "complete"
        ]
      }
    },
    "required": [
      "model",
      "usage",
      "status",
      "score",
      "level",
      "confidence",
      "probabilities",
      "completeness"
    ],
    "additionalProperties": false
  }
}

Saved example result

This hand-authored response demonstrates the contract. It is not a model accuracy measurement. Run it without an API key: npx jev-recipes demo handoff-completeness.

{
  "model": "demo-fixture",
  "usage": {
    "input_tokens": 0,
    "output_tokens": 0
  },
  "status": "ready",
  "score": 2.96,
  "level": 3,
  "confidence": 0.82,
  "probabilities": {
    "0": 0,
    "1": 0.02,
    "2": 0.08,
    "3": 0.82,
    "4": 0.08
  },
  "completeness": "defined"
}

Evaluation evidence

Use the evaluation guide to measure this decision on your own labeled cases.

Limitations

• Grades what the item states, not whether the stated goal is correct, feasible, or worth doing.
• Cannot know what the recipient already knows, so shared context that is not written into the item counts as missing.
• The score is an expected value over rubric levels. Application code chooses the level at which to hand off.

Related recipes

• plan-completeness: Use plan-completeness to grade whether a plan covers its goal, rather than whether a single work item is ready to hand off.
• clarify: Use clarify to find the specific missing requirements once an item grades below complete.