Research preview · 0.1.0-draft

No compatibility guarantee. Not for consequential production decisions.

Judgment Pack Specification
Synthetic example View tagged source

Expense approval

Research example for deciding whether an employee expense may be automatically approved.

Synthetic example: This document is for schema and authoring tests. It is not a policy, authorization, or instruction to perform an external action.
Pack ID
https://example.com/judgment-packs/expense-approval
Pack version
0.1.0
Outcomes
3
Rules
3

Download raw JSON

Guide to this example

Focus: Cross-feature authoring and local-reference tracing.

The checked-in document is expected to pass JPS carrier, structural, and semantic document checks. The edge and failure notes below are inspection prompts or scratch-copy exercises; they do not assert a portable evaluator result.

What this example demonstrates

A fuller JPS document with applicability, two evidence requirements, a cited source, three outcomes and rules, an exception, a fallback, and explicit escalation metadata.

Good for

  • Tracing outcome, evidence, and source IDs from declarations to every local reference.
  • Inspecting nested all/not conditions and the boundary between schema checks and semantic checks.
  • Reviewing how one synthetic pack records applicability, exceptions, fallback, and human-handoff configuration.

Edges to inspect

  • The declared amount split includes 5000 in the less-than-or-equal branch and values above 5000 in the review branch. Ordered decimal evaluation is still informative in this draft, so this is not a portable execution test.
  • In the informative model, an active-investigation forced outcome bypasses normal rules. Separately, a prohibited category plus an amount above 5000 can produce conflicting normal-rule candidates. Neither behavior is a portable evaluator result.
  • If activeInvestigation is absent, its exception is unknown and declares escalation in the informative model. Missing required evidence also requests handoff, while a non-employee-expense value is not applicable.
  • If resolution otherwise completes with no matching rule, the declared manual-review fallback applies. The listed no-match escalation trigger is unreachable in the informative algorithm while that fallback exists.

Failure paths

  • In a scratch copy, remove an outcome or evidence declaration while leaving a reference to it; the document can retain a valid JSON shape but must fail semantic document conformance.
  • Add an unknown root member to produce a structural failure, or duplicate a JSON member to produce a carrier failure before schema validation.
  • Never treat an illustrative approve, reject, or review label as authorization to process a real expense.
Keep the layers separate. A malformed carrier, a schema-shape failure, and a dangling local reference fail at different document-conformance layers. Missing evidence, unknown facts, conflicts, and no-match handling belong to the informative resolution experiment because JPS 0.1.0-draft defines no evaluator conformance class.

Annotated document

The complete validated pack. Each highlighted key links to its definition on the right — click a key to see what it means and its allowed values.

{
  "specVersion": "0.1.0-draft",
  "id": "https://example.com/judgment-packs/expense-approval",
  "version": "0.1.0",
  "title": "Expense approval",
  "description": "Research example for deciding whether an employee expense may be automatically approved.",
  "decision": {
    "intent": "Determine whether an employee expense may be automatically approved, rejected, or sent to a human reviewer.",
    "question": "What approval path should this expense follow?"
  },
  "applicability": {
    "op": "fact",
    "path": "/expense/type",
    "operator": "equals",
    "value": "employee-expense"
  },
  "evidenceRequirements": [
    {
      "id": "receipt",
      "description": "An itemized receipt for the expense.",
      "required": true,
      "kind": "document"
    },
    {
      "id": "cost-center",
      "description": "A valid cost-center identifier.",
      "required": true,
      "kind": "fact"
    }
  ],
  "sources": [
    {
      "id": "expense-policy",
      "title": "Example expense policy",
      "publisher": "Example Corporation",
      "publishedAt": "2026-01-15",
      "locator": {
        "kind": "uri",
        "value": "https://example.com/policies/expenses"
      },
      "citation": {
        "location": "Section 4 \u2014 approval thresholds",
        "excerpt": "Expenses above 5000 require finance review. Prohibited categories must be rejected."
      },
      "rights": "Example content; not an operational policy."
    }
  ],
  "outcomes": [
    {
      "id": "approve",
      "label": "Approve",
      "description": "The evidence and declared rules support automatic approval."
    },
    {
      "id": "manual-review",
      "label": "Manual review",
      "description": "An accountable finance reviewer must decide."
    },
    {
      "id": "reject",
      "label": "Reject",
      "description": "The expense falls in a prohibited category."
    }
  ],
  "rules": [
    {
      "id": "prohibited-category",
      "description": "Reject expenses in a prohibited category.",
      "when": {
        "op": "fact",
        "path": "/expense/category",
        "operator": "in",
        "value": [
          "personal",
          "cash-equivalent"
        ]
      },
      "outcome": "reject",
      "onUnknown": "escalate",
      "evidenceRequirementRefs": [
        "receipt",
        "cost-center"
      ],
      "sourceRefs": [
        "expense-policy"
      ],
      "rationale": "The example policy prohibits these categories."
    },
    {
      "id": "large-expense",
      "description": "Send expenses above the declared threshold to finance review.",
      "when": {
        "op": "fact",
        "path": "/expense/amount",
        "operator": "greater-than",
        "value": "5000"
      },
      "outcome": "manual-review",
      "onUnknown": "escalate",
      "evidenceRequirementRefs": [
        "receipt",
        "cost-center"
      ],
      "sourceRefs": [
        "expense-policy"
      ],
      "rationale": "The example policy requires finance review above 5000."
    },
    {
      "id": "ordinary-expense",
      "description": "Approve a supported ordinary expense at or below the threshold.",
      "when": {
        "op": "all",
        "conditions": [
          {
            "op": "fact",
            "path": "/expense/amount",
            "operator": "less-than-or-equal",
            "value": "5000"
          },
          {
            "op": "not",
            "condition": {
              "op": "fact",
              "path": "/expense/category",
              "operator": "in",
              "value": [
                "personal",
                "cash-equivalent"
              ]
            }
          },
          {
            "op": "evidence-present",
            "evidenceRequirement": "receipt"
          },
          {
            "op": "evidence-present",
            "evidenceRequirement": "cost-center"
          }
        ]
      },
      "outcome": "approve",
      "onUnknown": "escalate",
      "evidenceRequirementRefs": [
        "receipt",
        "cost-center"
      ],
      "sourceRefs": [
        "expense-policy"
      ],
      "rationale": "Ordinary supported expenses may follow the automatic path."
    }
  ],
  "exceptions": [
    {
      "id": "active-investigation",
      "description": "An expense connected to an active investigation requires human review.",
      "when": {
        "op": "fact",
        "path": "/expense/activeInvestigation",
        "operator": "equals",
        "value": true
      },
      "effect": "force-outcome",
      "outcome": "manual-review",
      "onUnknown": "escalate"
    }
  ],
  "fallbackOutcome": "manual-review",
  "escalation": {
    "triggers": [
      "not-applicable",
      "missing-required-evidence",
      "unknown",
      "conflict",
      "no-match"
    ],
    "target": {
      "kind": "human-role",
      "name": "Finance reviewer"
    },
    "message": "Review the expense, cited policy, missing evidence, and conflicting rules."
  },
  "metadata": {
    "authors": [
      "Judgment Pack Specification contributors"
    ],
    "createdAt": "2026-07-22T00:00:00Z",
    "license": "Apache-2.0"
  }
}